<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Bisma Hanif on Medium]]></title>
        <description><![CDATA[Stories by Bisma Hanif on Medium]]></description>
        <link>https://medium.com/@bismahanif88?source=rss-d60f4042ba49------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*bPyOo18_M6sbGNtjwUmoUQ.jpeg</url>
            <title>Stories by Bisma Hanif on Medium</title>
            <link>https://medium.com/@bismahanif88?source=rss-d60f4042ba49------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:36:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bismahanif88/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[What 90 Days of SEO Looks Like With 0 New Pages]]></title>
            <link>https://medium.com/@bismahanif88/what-90-days-of-seo-looks-like-with-0-new-pages-a322cf87765b?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/a322cf87765b</guid>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 07:41:58 GMT</pubDate>
            <atom:updated>2026-11-17T07:44:35.934Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-YfkViF1xeJBTMSH" /></figure><p><em>Two years of publishing produced 1,000 visits a month. Ninety days of fixing 60 existing pages produced 120,000.</em></p><p>I came across a case study this week that I have not been able to stop thinking about.</p><p>A B2B SaaS company spent two years with an agency. They published over 300 pages and spent tens of thousands of dollars a month doing it.</p><p>At the end of those two years they had under 1,000 monthly organic visits. The number was falling.</p><p>A new team came in and spent 90 days doing something different. They published nothing.</p><p><strong>Six months later the site was doing over 120,000 monthly organic visits.</strong> They just fixed what was already there.</p><p>I am not going to pretend every website has a 120x hiding in it. But the principle underneath that result is one of the most reliable things in SEO, and almost nobody acts on it.</p><h3>Why Everyone Keeps Publishing Instead</h3><p>Publishing feels like progress. It is countable, schedulable, and easy to put on a slide.</p><p>An agency can show you twelve new posts this month. It is much harder to sell a month spent improving things that already exist.</p><p>So the treadmill continues and the archive quietly rots behind it.</p><p>Meanwhile the odds on each new post are genuinely poor.</p><p>Ahrefs studied two million pages and found <strong>95% of newly published content never reaches the top 10 within its first year.</strong> Their analysis of pages currently sitting in the top 10 found the average result is over two years old.</p><p>So every new post is largely a lottery ticket, while the pages that already have history, links and ranking data sit untouched.</p><h3>Content Decay Is Real and It Is Measurable</h3><p>Here is the part most teams have never looked at.</p><p>Content does not hold its position by default. It decays, and it does so on a fairly predictable schedule.</p><p>Semrush content audit data found <strong>82% of high-ranking posts begin losing traffic within 12 to 24 months</strong> if nobody maintains them. ( <a href="https://thestacc.com/blog/content-update-impact-study/">Content Update Impact Study, 2026</a>)</p><p>Competitors publish something better. Statistics go stale. Search intent shifts. Your internal links point at pages you have since changed.</p><p>None of that shows up as a dramatic drop. It shows up as a slow slide that nobody attributes to anything in particular.</p><p>And the longer you leave it, the more of your archive is underperforming at once.</p><h3>What the Data Says About Fixing It Instead</h3><p>The returns on refresh work are better documented than almost anything else in content marketing.</p><p>HubSpot ran a formal historical optimisation programme and found that updating and republishing old posts produced a <strong>median 106% increase in organic traffic</strong> on those pages. Lead generation from the same posts more than doubled.</p><p>The more revealing HubSpot number is internal. 76% of their monthly blog views and 92% of their monthly blog leads come from old posts, not recent ones.</p><p>Read that second figure again. Nine in ten leads from content published in previous years.</p><p>Backlinko documented a <strong>260.7% organic traffic increase within 14 days</strong> after a strategic relaunch of an existing post. Worth saying clearly: that is a single page case study, not an average, and the result came from picking a page with genuine untapped demand.</p><p>Across broader samples, republished pages jump an average of 4.6 SERP positions. Moving from position 8 to position 3 is roughly a three to four times increase in click through rate.</p><p>A Databox survey found close to 90% of marketers say updating and repurposing existing content works better than producing new content from scratch.</p><p>Orbit Media’s annual blogger survey found that bloggers who update old posts are 2.5 times more likely to report strong results than those who do not.</p><p>Almost everyone agrees with this. Very few organisations actually staff it.</p><h3>The Audit: What You Are Actually Looking For</h3><p>A content audit is not a spreadsheet of every URL with its traffic next to it. That is an inventory.</p><p>An audit sorts your archive into buckets, each with a different action attached.</p><ul><li><strong>Pages close to ranking.</strong>Anything sitting in positions 5 to 20. These have already proven they can rank and need the smallest push for the largest return. Start here, always.</li><li><strong>Pages competing with each other.</strong>Two or more posts targeting the same query. Google splits confidence between them and often ranks none of them well.</li><li><strong>Pages that do not match intent.</strong>You wrote a how-to guide and the SERP is full of comparison pages. The content is fine. It is answering a different question than the one being asked.</li><li><strong>Pages with untapped potential.</strong>Ranking for a handful of terms when the topic supports dozens. Usually a coverage problem rather than a quality one.</li><li><strong>Pages that earn nothing.</strong>No traffic, no conversions, no links, no strategic purpose. These are candidates for consolidation or removal.</li></ul><p>Then prioritise ruthlessly. Not by what is worst, but by where the gap between current and achievable is largest.</p><h3>What to Actually Change</h3><p>Once you have the shortlist, the work itself is unglamorous and fairly mechanical.</p><ul><li><strong>Rewrite the introduction.</strong>Most old posts open with throat clearing. Answer the query in the first two sentences instead.</li><li><strong>Fix titles and metadata.</strong>The cheapest lever on this list. Question based title tags have been found to lift click through rate by around 14% on average.</li><li><strong>Strengthen internal linking.</strong>Point your strongest pages at the ones you are trying to lift. Most sites have authority sitting in pages that link nowhere useful.</li><li><strong>Improve structure.</strong>Clear headings, self contained answers, scannable formatting. This helps readers and it is also how AI systems extract passages.</li><li><strong>Expand topical coverage.</strong>Add the subtopics the page never covered. Depth beats length, though the two often arrive together.</li><li><strong>Rematch search intent.</strong>Check what is actually ranking for your target query and whether your format matches it. Sometimes the fix is changing the page type entirely.</li><li><strong>Consolidate overlapping posts.</strong>Merge the competing pages into one authoritative version and redirect the rest. This is usually the single biggest win in the whole audit.</li><li><strong>Update the facts.</strong>Stale statistics and outdated screenshots are a credibility problem for readers and a freshness problem for search engines.</li></ul><h3>Do Not Skip the Deletions</h3><p>The uncomfortable part of this work is removal, and the evidence for it is strong.</p><p><strong><em>QuickBooks deleted 2,000 blog posts, more than 40% of their Resource Center. Traffic rose 20% within weeks and 44% by peak season, and sign ups increased 72%.</em></strong></p><p>Inflow pruned roughly 200 pages from an e-commerce client’s blog and strategic content revenue increased 64%. HubSpot removed around 3,000 old posts and saw traffic rise afterwards.</p><p>Thin pages do not just fail on their own. They pull down the average quality signal across your whole domain.</p><p>If a page has no traffic, no conversions, no links and no strategic reason to exist, it is not neutral. It is costing you.</p><h3>This Matters More in the AI Era</h3><p>One argument I hear is that refresh work belongs to the old search world. The opposite is true.</p><p>AI answer engines lean heavily on recently updated pages. Freshness is a stronger signal in citation than it ever was in ranking.</p><p>A page last touched in 2023 is competing for citations against pages updated this quarter, on queries where the model is explicitly looking for current information.</p><p>Refreshed content also converts better. One analysis found strategically refreshed pages convert around 27% better than new content published on the same topic.</p><p>Which makes sense. A refreshed page carries its existing authority, its backlink history and its ranking data. A new page starts from nothing.</p><h3>How to Run This on Your Own Site</h3><p>If you want to test the principle before committing a quarter to it, here is the smallest useful version.</p><ul><li><strong>Pull every URL with its clicks, impressions and average position</strong>from Search Console, for the last 12 months.</li><li><strong>Filter to positions 5 to 20.</strong>These are your fastest wins. Pages already close enough that a real improvement moves them.</li><li><strong>Pick ten.</strong>Not sixty. Ten, so you can do them properly and measure what happened.</li><li><strong>Check each one against the current SERP.</strong>What is ranking, what format is it, and what does it cover that you do not?</li><li><strong>Rewrite rather than tinker.</strong>A few new sentences will not move anything. Treat it as a new draft that keeps the URL.</li><li><strong>Wait 30 to 60 days, then measure.</strong>Most well diagnosed refreshes show ranking movement inside a month.</li></ul><p>If ten pages move, you have evidence to fund the other fifty. If they do not, you have learned something cheap.</p><h3>So, What’s My Take?</h3><p>The publishing treadmill persists because it is easy to sell and easy to measure. Twelve posts a month looks like work. A quarter spent fixing the archive looks like admin.</p><p>But the data has pointed the same direction for years. Old pages carry the authority. Old pages generate most of the leads. New pages mostly do not rank.</p><p>That SaaS company spent two years and a very large budget publishing 300 pages to reach 1,000 visits a month.</p><p>The fix took 90 days and produced 120,000.</p><p><strong>Your archive is not finished work. It is an asset you stopped maintaining.</strong></p><p>Before you commission the next twelve posts, go and look at what you already paid for.</p><p><em>Originally published at </em><a href="http://bismamenon.wordpress.com/2026/11/17/what-90-days-of-seo-looks-like-with-0-new-pages/"><em>http://bismamenon.wordpress.com</em></a><em> on November 17, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a322cf87765b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Difference Between SEO and GEO Is Simpler Than You Think]]></title>
            <link>https://medium.com/@bismahanif88/the-difference-between-seo-and-geo-is-simpler-than-you-think-ac807b2ff088?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/ac807b2ff088</guid>
            <category><![CDATA[seo-tips]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[ai-seo]]></category>
            <category><![CDATA[seo]]></category>
            <category><![CDATA[seo-strategy]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Fri, 04 Sep 2026 16:39:54 GMT</pubDate>
            <atom:updated>2026-09-04T16:43:55.743Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AfgmscuEivOY273b" /></figure><p>People ask me this constantly. How is GEO actually different from SEO? What would you do differently?</p><p>The answer traces back to something very basic, and once you see it, every strategy decision downstream becomes obvious.</p><ul><li><strong>Search engines show what you are saying about yourself.</strong></li><li><strong>Generative engines show what others are saying about you.</strong></li></ul><p>That is the whole difference. Everything else is a consequence of it.</p><h3>1. So What Would You Do on Your Own Website to Get Cited in AI Search?</h3><p>This is the question I get asked most, and it is the most misunderstood one in the field right now.</p><p>Before answering it, try something. It takes five minutes and it will change how you think about this.</p><p>Search your own brand in ChatGPT, Gemini, or Google AI Overviews. Ask a few different types of questions about your category.</p><p>Then look at the citations underneath the answer.</p><p><strong>How many of them point to your own website? And for which questions specifically?</strong></p><p>For most brands, the answer is: very few, and only for a narrow band of queries.</p><h3>The numbers behind that</h3><p>This is one of the most consistently measured findings in the whole field.</p><p>Muck Rack analysed more than 25 million links cited by ChatGPT, Claude and Gemini across 17 industries. <strong>Earned third-party media accounted for 84% of all AI citations.</strong> That figure has held between 82% and 89% across three editions of the study. ( <a href="https://www.cognizo.ai/blog/third-party-ai-citations">Cognizo</a>)</p><p>An academic analysis of 131,514 citations found <strong>85.7% pointed to sites the brand does not own, against 14.3% owned.</strong> The authors summarised it neatly: AI reads third parties about a brand four to six times as often as it reads the brand’s own site. ( <a href="https://arxiv.org/pdf/2606.25787">arXiv</a>)</p><p>AirOps’ 2026 State of AI Search report landed in the same place, at 85% of brand mentions originating from third-party pages.</p><p><strong>Interestingly, B2C brands are slightly more self-cited than B2B, at 15.7% against 13.1%. Self-citation does not vanish. It concentrates.</strong></p><h3>2. Which Queries Cite Your Site, and Which Cite Everyone Else</h3><p>Here is the part that makes this practical, because the split is not random. It is predictable.</p><p>Your own website gets cited when the question is a <strong>bout facts only you can supply</strong>.</p><ul><li>What your product does.</li><li>What it costs.</li><li>What is included in which plan.</li><li>Where your offices are.</li><li>Who your leadership is.</li><li>How to contact support.</li></ul><p>Trustmary’s analysis found brand-owned domains account for roughly <strong>62% of citations on brand-specific queries</strong> where someone asks about your company by name. So when the user already knows who you are, you are the source.</p><p>Now change the question.</p><p><em>Ask an AI whether Product X is better than Product Y. Ask for the best tool in a category. Ask what the drawbacks of a product are.</em> <strong>Your website will almost never be quoted.</strong></p><ul><li>Sometimes it cites your competitors.</li><li>Sometimes a review platform.</li><li>Sometimes a third-party listicle.</li><li>Sometimes a YouTube or TikTok video reviewing your product alongside a hundred others.</li></ul><p>It’s simply the editorial logic that these systems were trained into, which is <strong><em>citing the most authentic answer, not self-promotion!</em></strong></p><p>A page where you describe your product is useful, but it is not independent evidence.</p><p>When only the brand makes a claim, the model treats it as promotional. When several unrelated sources say the same thing, it treats it as fact.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/861/1*c3kIpnN0XVutDGlrugWTfQ.jpeg" /></figure><p>Look at that table again. Every row in the top half is someone who already knows you. Every row in the bottom half is someone deciding whether to buy.</p><p><strong>Your own site covers the queries from people already sold. Third parties cover the queries that decide the sale.</strong></p><p>A Conductor study running over seven months found the same pattern from another angle.</p><p><strong>Google AI Overviews preferred YouTube as the top-cited domain</strong> for five of seven intent types. The single exception was navigational intent, where brand-owned domains took the top slot.</p><h3>3. Where the Citations Actually Come From</h3><p>If most of your visibility sits off your domain, the obvious question is where.</p><p>The concentration here is more extreme than anything PageRank ever produced.</p><p>Peec AI analysed 30 million directly cited sources across ChatGPT, Google AI Mode, Gemini, Perplexity and AI Overviews.</p><p><strong>Reddit ranked as the single most cited domain on all five.</strong></p><p>Consolidated index of over 680 million citations found the top 15 domains capture roughly 68% of everything, with Reddit alone at around 40%. ( <a href="https://searchengineland.com/ai-search-engines-cite-reddit-youtube-and-linkedin-most-study-473138">Search Engine Land</a>)</p><p>The rest of the list is predictable once you see it. YouTube, LinkedIn, Wikipedia, Forbes, TechRadar. For commercial queries, G2 and Capterra. For consumer categories, Trustpilot and review aggregators.</p><p>SE Ranking studied 129,000 domains and found brands with a substantial Reddit presence averaged <strong>7 ChatGPT citations, against 1.8 for brands with minimal presence.</strong> Almost four times the visibility.</p><p>And one detail that should reframe your whole measurement approach.</p><p>Moz analysed 40,000 queries and found <strong>88% of Google AI Mode citations do not appear in the organic top 10.</strong> A separate study found 37% of AI-cited domains are entirely absent from traditional search results.</p><p>Your rank tracker is not measuring this. It was never built to.</p><h3>4. So What Is the Actual Strategy?</h3><p>Here is how I would allocate effort, in order of return.</p><h3>Platforms to prioritize for outreach</h3><ul><li><strong>Reddit, first and by a distance.</strong> Around 40% of all citations. But participate honestly or not at all. These systems cite Reddit precisely because it reads as unfiltered opinion. Turn up as an advert and you will be removed, deservedly.</li><li><strong>Review platforms in your category.</strong> G2 and Capterra for software, Trustpilot for consumers, Yelp for local. These are the sources cited on exactly the queries that decide purchases.</li><li><strong>YouTube.</strong> Top-cited domain for five of seven intents in AI Overviews. Note that AI does not watch the video: it reads the transcript, captions, chapters and description. Video SEO is text SEO in a different outfit.</li><li><strong>LinkedIn.</strong> The most cited domain for professional and category comparison queries, which is precisely the B2B buying question.</li><li><strong>Wikipedia and reference sources.</strong> Entity clarity is a strong citation signal. If you are not a documented entity anywhere neutral, models have less to anchor to.</li><li><strong>Category listicles and comparison articles.</strong> The “best X for Y” pages already being cited in your category. Getting included is worth more than publishing your own version.</li></ul><h3>Is PR worth it? Yes, and there is a controlled study</h3><p>This is normally where the conversation gets vague. It does not have to be.</p><p>Stacker and Scrunch ran a controlled test: 8 articles across 944 prompt and platform combinations on five leading LLMs. <strong>Content on a brand’s own site was cited 8% of the time. The same content distributed through third-party news outlets was cited 34% of the time.</strong> A 325% lift, same words. ( <a href="https://authoritytech.io/blog/machine-relations-evidence-earned-media-ai-citations">AuthorityTech</a>)</p><p>A follow-up in March 2026 across 87 stories and 30 brands found a 239% median lift.</p><p>The same research found that cited press releases had 30% more objective sentences and 2.5 times as many bullet points as uncited ones. Write for extraction, not persuasion.</p><p>One caveat worth knowing: paid and advertorial content accounted for just 0.3% of citations in the Muck Rack data. Buying placement does not work. Earning it does.</p><h3>What actually predicts citations</h3><p>Ahrefs studied 75,000 brands and compared signals.</p><p>Brand mentions across independent sources correlated <strong>0.664 with AI citation probability. Backlinks correlated just 0.218.</strong></p><p>Who talks about you matters roughly three times more than who links to you.</p><p>That single finding should redirect a meaningful share of most link building budgets.</p><h3>5. What to Change on Your Own Website</h3><p>Your site still matters. It just has a narrower and more specific job now.</p><p>It is the authoritative source for facts about you, and it is what models check when they need to verify a claim someone else made.</p><ul><li><strong>Make pricing public and machine readable.</strong> Pricing is one of the highest intent queries you own. If it is behind a form, the model cites a third party guessing at it.</li><li><strong>Write clear product and comparison pages.</strong> Including honest comparisons against competitors. Competitor product pages are already among the highest cited domains in most categories. Self-promotion doesn’t work here at all.</li><li><strong>Use structured data properly.</strong> Organization, Product, FAQ and Review schema. This is how you become a legible entity rather than prose to be parsed.</li><li><strong>Answer questions in self-contained blocks.</strong> A question as a heading, a complete answer directly underneath. Models extract passages, not pages.</li><li><strong>Keep facts consistent everywhere.</strong> Your name, category description, founding details and product claims should match across your site, LinkedIn, Wikipedia and directories. Contradictions reduce confidence.</li><li><strong>Publish original data.</strong> Proprietary research is the one type of owned content that gets cited widely, because nobody else has it.</li></ul><p><strong>But be realistic about the ceiling. On-site work optimizes roughly 15% of your AI visibility.</strong> Doing it perfectly will not move the other 85%.</p><p>The most common mistake I see in 2026 is teams spending their entire GEO budget on their own website, then wondering why nothing changed.</p><h3>6. Are Fan-Out Queries Worth Writing For?</h3><p>Short answer: understand them, but avoid chasing them individually.</p><p><strong><em>Fan-out is how these systems actually work. When you ask a question, the model breaks it into sub-questions, runs them in parallel, and assembles an answer from whatever each one returns.</em></strong></p><p>Google’s own documentation describes it, and the scale is significant. Google indicated the average AI Mode query generates 12 to 15 sub-queries, with complex ones expanding past 50.</p><p>So a page that answers more of those sub-questions is more likely to be pulled into the answer, even if it does not rank first for anything.</p><h3>Why chasing individual fan-outs fails</h3><p>Two findings make this clear.</p><ul><li><strong>Only around 27% of sub-queries stay stable</strong> across repeated searches. ChatGPT, AI Mode and Perplexity all generate different ones. You would be optimising for a target that changes between runs.</li><li><strong>Roughly 95% of fan-out phrases show zero monthly search volume.</strong> They will never appear in a keyword tool, so you cannot build a normal content plan around them anyway.</li></ul><p>Instead of one page per keyword, build genuinely comprehensive coverage of a topic: the definition, the comparison, the pricing question, the objection, the use case, the alternative.</p><p><strong>Fan-out queries average 5.5 words on ChatGPT and 9.1 on Gemini, against roughly 3.4 for a classic Google search</strong>. Longer, more specific, more conversational.</p><p>Write the way a knowledgeable person would actually answer a question, and you cover the fan-out by accident. That is the honest version of this advice.</p><h3>7. Putting It Together</h3><p>If I were rebuilding a visibility strategy from scratch today, this is the order.</p><ul><li><strong>Audit the split first.</strong> Run twenty real questions your buyers ask across ChatGPT, Gemini and AI Overviews. Record every citation. You now know where you actually stand.</li><li><strong>Fix the owned 15%.</strong> Pricing, product pages, schema, consistent entity facts. This is a few weeks of work, not a strategy.</li><li><strong>Move budget to earned.</strong> This is where the 84% lives, and where most teams currently spend almost nothing.</li><li><strong>Get into the sources already cited in your category.</strong> Not new sources. The ones appearing in your answers today.</li><li><strong>Build genuine community presence.</strong> Reddit and category forums, over months, honestly. There is no fast version of this.</li><li><strong>Track citations per platform, not in aggregate.</strong> Different models cite very different sources. Averaged reporting hides the actionable part.</li></ul><h3>So, What’s My Take?</h3><p>SEO asked one question: is your page good enough to rank?</p><p>GEO asks a different one: do enough independent sources say the same thing about you that a model treats it as fact?</p><p><strong>GEO is more of a reputation management with structured data attached.</strong></p><p>Which is why the teams struggling most with GEO are often the ones with the strongest technical SEO. They are extremely good at controlling their own pages, and this game is largely played on everyone else’s.</p><p><strong>You cannot optimize your way into being trusted. You can only be talked about accurately, in enough places, by enough people who do not work for you.</strong></p><p>Your website tells AI what you claim. Everything else tells it whether to believe you.</p><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/09/04/seo-vs-geo/"><em>http://bismamenon.wordpress.com</em></a><em> on September 4, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ac807b2ff088" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Google’s Recent Launch the Worst Monopoly Move Ever?]]></title>
            <link>https://medium.com/@bismahanif88/is-googles-recent-launch-the-worst-monopoly-move-ever-a25f9ec2cf5f?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/a25f9ec2cf5f</guid>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[website-traffic]]></category>
            <category><![CDATA[goto-redirect]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[google]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Wed, 02 Sep 2026 12:52:40 GMT</pubDate>
            <atom:updated>2026-09-02T12:56:11.424Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mvjwj3I_5lfxcjPs" /></figure><p>Google has spent the last two years quietly levelling up its monopoly game, and it has been very good at it.</p><p><strong>AI Overviews</strong> arrived and started answering the question on the results page, so nobody needed to click through. <strong>Gemini</strong> answers before a results page even appears.</p><p><strong>YouTube, Maps, Shopping and Flights</strong> hold the user inside Google properties for the whole journey. <strong><em>Zero click searches reached 68% by 2026.</em></strong></p><p>Every one of those moves kept more of the value inside Google. Publishers absorbed the cost.</p><p>And then, in August, Google did something new. Something that does not touch your website at all.</p><p>Which is precisely why it is worth paying attention to.</p><h3>What this new Launch is About?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vrxWXS72UO_ftT7z" /></figure><p>On 26 August 2026, Google confirmed it is rolling out something called <strong>goto redirects</strong> across search results.</p><p>Here is the plain English version.</p><p>Until now, when you searched for something, the link in the results pointed straight at the website. You could hover over it and read the address.</p><p>Now the link points at Google first. It looks like this: google.com/goto?url=[a long scrambled string].</p><p>You click it, Google receives that click, and Google forwards you on to the real website.</p><p><strong>Think of a phone directory that used to print the number, and now prints a code. To reach anyone you have to call the directory first and let them connect you.</strong></p><p>The call still goes through. The directory just now knows about every single one.</p><h3>The important detail</h3><p>That scrambled string cannot be unscrambled.</p><p>There is no way to look at the link and work out where it goes. The only way to find out is to actually click it and see where you land.</p><p>Hold onto that, because it is the whole point of the exercise.</p><h3>Why This Does Not Affect Normal Websites</h3><p>Let me get the reassuring part out of the way, because a lot of the coverage has been unnecessarily alarming.</p><p>If you run a website, this changes nothing for you.</p><ul><li><strong>Your rankings are untouched.</strong>This happens after Google has already decided where you appear. It is about the click path, not the ordering.</li><li><strong>Your traffic still arrives.</strong>Users land on exactly the same page they always did, one invisible step later.</li><li><strong>Search Console is unaffected.</strong>Your reporting still works the way it did last month.</li><li><strong>There is nothing to fix.</strong>No change to your URLs, canonical tags, redirects or site structure. Anyone selling you a goto audit is selling you nothing.</li><li><strong>Users will not notice.</strong>No warning screen, no extra page, no delay they will feel. Most people will never know this happened.</li></ul><p>Google has done this with <strong>ads for years</strong>. Paid results have run through a similar redirect for a long time and nobody lost sleep over it.</p><p>So if you are a business owner reading this: relax. This one is not aimed at you.</p><h3>So Who Is It Aimed At?</h3><p>Everybody whose job is to read Google’s results page without being a human.</p><p>That means <strong><em>rank tracking tools, SERP data providers, competitive intelligence platforms, scrapers, and the AI engines</em></strong> that pull search results to build their own answers.</p><p>All of those services work the same basic way. They load a results page, read the links, and record which websites appeared where.</p><p>That job just became dramatically more expensive.</p><h3>The maths of it</h3><p>Before, a tool could read a results page once and know every destination. One request, full picture.</p><p>Now, because the links cannot be decoded, the tool has to follow every single one to find out where it goes.</p><p>PPC Land reported that this pushes rank trackers from a single request into <strong>500 to 1,000 requests per query.</strong> ( <a href="https://searchengineland.com/google-confirms-deploying-goto-url-redirects-to-search-results-links-485926">Search Engine Land</a>)</p><p>Read that again. <strong>Up to a thousand times the work to learn what used to be free.</strong></p><p>For a large platform tracking millions of keywords daily, that is not an inconvenience. That is a business model problem.</p><p>Derek Perkins of Nozzle, one of the first people to document it, reported nearly 100% rollout across several residential IP providers.</p><h3>And the AI engines</h3><p>Here is the detail I find most telling.</p><p>Analysts noted that AI Overviews are currently the hardest part of the results page to resolve cleanly.</p><p>So the part of Google that competes most directly with independent AI answer engines is also the part that is hardest for outsiders to read.</p><p>That may be coincidence. It is a very convenient one.</p><h3>What Google Says About It</h3><p>This is the part that tells you the most, and it is worth quoting exactly.</p><p>A Google spokesperson said:</p><blockquote>“We have a long history of deploying technical measures against evolving forms of abuse, and we regularly take steps to protect our services and users.”</blockquote><p>That is the entire statement.</p><p>Google did not name scrapers. Did not name rank trackers. Did not name AI companies. Did not publish a blog post, documentation, or a Search Central entry.</p><p>No timeline, no scope, no explanation of what abuse means in this context.</p><p><strong>A change rolled out across essentially every search result in the world, explained in twenty three words.</strong></p><p>And the framing does a lot of quiet work. Calling it protection against abuse puts rank trackers and AI engines in the same sentence as bad actors, without ever having to say so directly.</p><h3>The context Google left out</h3><p>There is a piece of background that makes the timing more interesting.</p><p>Google has been in litigation with SerpApi, a search results scraping service, and recently lost a motion in that case.</p><p>Losing in court is slow, expensive, and public. Making the data expensive to collect is fast, cheap, and requires no permission from anybody.</p><p>You can decide for yourself which of those looks more appealing to a legal team.</p><h3>Is It the Worst Monopoly Move Ever?</h3><p>Since the title asks the question, let me answer it properly rather than pretending outrage I do not feel.</p><p>This is not the worst thing Google has done to the open web in the last two years. It is not close.</p><p>AI Overviews cut position one click through rate from 7.3% to roughly 1.6%. That decision took real revenue from real publishers, at scale, permanently.</p><p>The goto rollout, by comparison, makes life difficult for a few hundred software companies. Most of them have already adapted.</p><p>DataForSEO said it already resolves 99.99% of these URLs and shipped an update covering AI Overviews within two days. Semrush reported no impact on rank tracking or data accuracy.</p><p>So as a piece of sabotage, it has been fairly comfortably absorbed.</p><h3>But Here Is Why It Still Matters</h3><p>The size of the move is not the interesting part. The direction is.</p><p>Every previous change made Google more valuable to Google. This one makes Google harder to observe.</p><p>Those are different things, and the second one is more concerning over time.</p><p><strong>Rank tracking is how the industry checks Google’s work.</strong> It is how we spot volatility, verify algorithm updates, catch preferential treatment, and prove to a client that something changed.</p><p>It is also, quite often, how regulators and researchers understand what Google is actually doing.</p><p>Make that measurement expensive enough and it does not disappear, it just consolidates. Only the largest tools can afford to keep doing it.</p><p>Smaller platforms, independent researchers and academics get priced out first. Fewer people watching, and the ones left have the most to lose from a falling out.</p><p>None of that is illegal. Some of it is genuinely defensible as anti abuse work. All of it happens to make Google less accountable.</p><h3>So, What’s My Take?</h3><p>I do not think this is the worst thing Google has done. I think it is the most revealing.</p><p>The changes that hurt publishers were at least visible. You could measure the traffic drop, write about it, and argue about whether it was fair.</p><p>This one is invisible by design. It costs users nothing, costs websites nothing, and quietly raises the price of watching Google to somewhere between 500 and 1,000 times what it used to be.</p><p><strong>A company confident in its own fairness does not usually make itself harder to audit.</strong></p><p>For your business, do nothing. Genuinely. Check your reporting tool has adapted, and move on with your quarter.</p><p>But keep noticing the pattern. First they kept the clicks. Now they are making it harder to prove it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*5mKjOyn7JLrHnmna" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/09/02/google-goto-redirect/"><em>http://bismamenon.wordpress.com</em></a><em> on September 2, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a25f9ec2cf5f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[SEO Is Not Enough in 2026. You Need More Than That.]]></title>
            <link>https://medium.com/@bismahanif88/seo-is-not-enough-in-2026-you-need-more-than-that-bf3e47336e8c?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf3e47336e8c</guid>
            <category><![CDATA[conversion-optimization]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[search-engine-optimizaion]]></category>
            <category><![CDATA[lead-management]]></category>
            <category><![CDATA[seo-2026]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Wed, 19 Aug 2026 15:53:12 GMT</pubDate>
            <atom:updated>2026-08-19T15:58:24.877Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rCpggvZhwz6B5h0q" /></figure><p><em>S</em>EO<em> is not enough in AI world. Here is what a complete visibility system looks like now.</em></p><p>Every few months someone announces that SEO is dead. It is not, and that has never been the interesting question.</p><p>The interesting question is whether SEO on its own is still enough to grow a business.</p><p>It is not. And the reason is simple: ranking first only helps if your buyer is actually looking there.</p><p>Increasingly, they are looking in five other places first.</p><h3>SEO is Not Dead. It Fragmented.</h3><p>Ten years ago, search meant Google. One box, one ranking, one game.</p><p>Now the same person researching a purchase might open TikTok for inspiration, Reddit for honest opinions, YouTube for a demonstration, LinkedIn for credibility, and ChatGPT to summarise the options.</p><p>Google is still in that journey. It is just no longer the whole journey.</p><p>Adobe Express surveyed 807 consumers in January 2026 and found <strong>49% have used TikTok as a search engine,</strong> up from 41% in 2024. Among Gen Z that figure reaches 65%. ( <a href="https://www.searchenginejournal.com/gen-z-preference-for-tiktok-over-google-drops-50-data-shows/568267/">Search Engine Journal</a>)</p><p>A Forbes study found <strong>67% of Gen Z use Instagram for search, 62% use TikTok, and 61% use Google.</strong> Almost identical shares across all three.</p><p>Google’s own executives said years ago that roughly 40% of people aged 18 to 24 go to TikTok or Instagram, not Google, when looking for somewhere to eat.</p><h3>The idea most people get wrong</h3><p>Here is where a lot of commentary goes off the rails.</p><p>This is not substitution. It is distribution.</p><p>The same Adobe research found that Gen Z preference for TikTok <em>over</em> Google actually fell, from 8% in 2024 to 4% in 2026. And a Semrush cohort study found that adopting ChatGPT produced no statistically significant change in daily Google sessions.</p><p>People are not abandoning Google. They are adding platforms and choosing between them based on the question.</p><p>Someone searches TikTok for “best skincare routine for oily skin,” then twenty minutes later searches Google for “dermatologist near me.” Both behaviours belong to the same person.</p><p><strong>So the goal is not picking the winning platform. It is being present across the journey.</strong></p><h3>Where AI Actually Gets Its Answers</h3><p>If your buyers now ask ChatGPT or Perplexity who to trust, it is worth knowing what those systems read before answering.</p><p>The data here is unusually clear, and it should change how you allocate effort.</p><p>Peec AI analysed 30 million directly cited sources across ChatGPT, Google AI Mode, Gemini, Perplexity and AI Overviews in May 2026. <strong>Reddit ranked as the single most cited domain across all five engines.</strong> ( <a href="https://searchengineland.com/ai-search-engines-cite-reddit-youtube-and-linkedin-most-study-473138">Search Engine Land</a>)</p><p>YouTube, LinkedIn, Wikipedia and Forbes rounded out the top five. Review platforms like G2 and Yelp appeared heavily in recommendation queries.</p><p>A consolidated index of more than 680 million citations found the top 15 domains capture roughly 68% of all AI citations, with Reddit alone accounting for around 40%.</p><p>The concentration is more extreme than PageRank ever was.</p><h3>What that means practically</h3><ul><li><strong>Reddit presence has measurable citation value.</strong>SE Ranking studied 129,000 domains and found brands with a substantial Reddit presence averaged 7 ChatGPT citations, against 1.8 for brands with minimal presence. A 3.9x difference.</li><li><strong>YouTube has become a citation engine.</strong>Goodie AI tracked YouTube’s social citation share doubling from 18.9% to 39.2% between August and December 2025.</li><li><strong>LinkedIn owns professional queries.</strong>Profound analysed 1.4 million citations and found LinkedIn the most cited domain for professional and category comparison questions, exactly the shape of query a B2B buyer asks.</li><li><strong>AI does not watch your video.</strong>It parses the transcript, captions, chapter markers, title and description. Video SEO is text SEO wearing a different outfit.</li></ul><p>None of that is on your website. All of it decides whether you show up in the answer.</p><h3>Search Everywhere Optimization</h3><p>The strategy that follows from all this is simpler than it sounds, and it does not require five separate content teams.</p><p>Answer one question your buyer is genuinely asking. Then publish that answer in every place they might ask it.</p><p>One properly researched answer becomes a YouTube video. That video becomes shorts and reels. The transcript becomes a blog post. The core argument becomes a LinkedIn post. The honest version becomes a Reddit comment in a thread where someone asked.</p><p>Same insight. Six placements. One research effort.</p><p>One caution on Reddit. Participate honestly or do not participate at all.</p><p>These systems cite Reddit precisely because it reads as unfiltered opinion rather than marketing. Turn up as an advert and you will be removed, and you will deserve it.</p><h3>But Being Found Is Only Half the System</h3><p>Here is the part that most visibility advice skips entirely.</p><p>Being seen does not make money. Converting makes money.</p><p>I have watched businesses build genuinely impressive reach across five platforms, celebrate the follower graph, and generate almost no revenue from any of it.</p><p>Visibility without a conversion path is a hobby with good analytics.</p><h3>Put budget behind what already works</h3><p>Most teams create new content and hope something lands. The better move is the opposite.</p><p>Find the post, video or page that already performed organically. That content has proven demand and proven messaging.</p><p>Then put a modest paid budget behind it. You are not gambling on a creative idea; you are buying more distribution for something the audience already validated.</p><h3>Send everything to one clear offer</h3><p>Fragmented content is fine. A fragmented destination is not.</p><p>If every piece of content points somewhere different, you are teaching people that you do many things and giving them no reason to act on any of them.</p><p>Pick the single next step you want a stranger to take. A free trial, a demo, a template, a tool. Then point everything at it.</p><h3>Capture the lead and follow up automatically</h3><p>This is where most of the money is lost, and it is entirely fixable.</p><p>Every lead should land in a CRM with follow up already built in. Email sequences, and text where the category justifies it.</p><p>The economics are not close. Email returns <strong>$36 to $42 for every dollar spent in 2026,</strong> against $2.00 for paid search and $2.80 for social advertising.</p><p>Klaviyo’s data found automated flows generated around 41% of email revenue from just 5.3% of sends. The follow up sequence does the work, not the volume.</p><p>And this is the only part of the entire system that cannot be taken away by an algorithm change.</p><h3>What a Complete 2026 System Looks Like</h3><p>Put together, the pieces are straightforward.</p><ul><li><strong>Answer real questions, not keyword phrases.</strong>Start from what your buyer actually types or says, on whichever platform they say it.</li><li><strong>Publish everywhere they search.</strong>One answer, adapted per format. Video, short, article, professional post, community reply.</li><li><strong>Be present in the sources AI reads.</strong>Reddit, YouTube and LinkedIn are not social channels any more. They are citation infrastructure.</li><li><strong>Amplify what already works.</strong>Small paid budgets behind proven organic content beat large budgets behind guesses.</li><li><strong>Drive everything to one offer.</strong>One clear next step, repeated everywhere.</li><li><strong>Capture and nurture automatically.</strong>CRM, sequences, and a list you own outright.</li><li><strong>Measure conversion, not reach.</strong>Impressions are the vanity half. Pipeline is the half that pays.</li></ul><h3>So, What’s My Take?</h3><p>SEO is not dead and I am not going to pretend it is. It is still the best acquisition channel most businesses have, and it still compounds better than anything else.</p><p>What changed is that it stopped being sufficient on its own.</p><p>Your buyers are asking their question in five places and trusting whichever answer they find first. If you are only present in one of those places, you are competing for a shrinking share of a single channel.</p><p>And if you win that visibility but have no conversion system behind it, you have built an audience for someone else to sell to.</p><p><strong>Getting found is the first half. Turning that into a relationship you own is the half that actually pays.</strong></p><p>Do both, and the algorithm changes stop being existential.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*ozQyIXKQLNMQEd00" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/19/seo-is-not-enough-in-2026-you-need-more-than-that/"><em>http://bismamenon.wordpress.com</em></a><em> on August 19, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf3e47336e8c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[You Are Renting Your Traffic From Google. Here’s How to Own It]]></title>
            <link>https://medium.com/@bismahanif88/you-are-renting-your-traffic-from-google-heres-how-to-own-it-c4e2e2c1c322?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/c4e2e2c1c322</guid>
            <category><![CDATA[lead-generation]]></category>
            <category><![CDATA[email-marketing]]></category>
            <category><![CDATA[conversion-optimization]]></category>
            <category><![CDATA[ai-seo]]></category>
            <category><![CDATA[seo-strategy]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Sun, 16 Aug 2026 15:39:58 GMT</pubDate>
            <atom:updated>2026-08-16T15:42:01.988Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*O2yXEphEEajwCSol" /></figure><p><em>SEO alone is no longer a growth strategy. It is a tenancy agreement, and the landlord keeps raising the rent.</em></p><p>Here is an uncomfortable question. If <strong>Google turned off tomorrow</strong>, how many people could you still reach?</p><p>For most businesses the honest answer is: <em>almost none</em>.</p><p>That is a huge business risk, and it has been sitting quietly on the balance sheet for years.</p><p>You do not own your traffic. You rent it. And the landlord has started renovating.</p><h3>The Landlord Is Changing the Terms</h3><p>Google is no longer a place that sends people away. It is increasingly a destination that keeps them.</p><p><strong>AI Overviews </strong>answer the question on the results page. <strong>Gemini</strong> answers it before a results page appears. YouTube, Maps, Shopping, and Flights hold the user inside Google properties for the whole journey.</p><p>None of this is a conspiracy. It is a rational business decision. Google earns more when attention stays on Google.</p><p>The problem is that an entire industry built its growth model on the assumption that Google would keep handing that attention out.</p><p><strong>What the Numbers Actually Look Like</strong></p><p>This is where it stops being a theory.</p><p>Pew Research tracked 68,879 real Google searches. When an AI Overview appeared, users clicked a traditional result <strong>8% of the time, against 15% without one.</strong> Around 1% clicked a link inside the summary. ( <a href="https://www.similarweb.com/blog/marketing/geo/zero-click-marketing/">Pew via Similarweb</a>)</p><p>Ahrefs studied 300,000 keywords. Position one click through rate fell to roughly <strong>1.6% with an AI Overview present, down from 7.3%</strong> two years earlier.</p><p>The detail that should worry you most: even without an AI Overview, position one fell to 3.9%. Ranking first is simply worth less than it used to be.</p><p>SparkToro’s 2026 clickstream study found zero click searches reached <strong>68%,</strong> with the share of searches producing any click down 22.9% between 2024 and 2026. ( <a href="https://searchengineland.com/google-zero-click-searches-2026-study-479717">Search Engine Land</a>)</p><p>And ranking is not stable either. In the March 2026 core update, 79.5% of top three URLs changed position and 24.1% of top ten pages fell out of the top 100 entirely.</p><p>You can do everything right and still lose the tenancy.</p><h3>Why This Is a Risk, Not Just a Dip</h3><p>If your entire funnel starts with an organic click, then every part of your business depends on a system you do not control, cannot audit, and did not agree terms with.</p><p>One algorithm update can remove a quarter of your pipeline in a week. There is no notice period and no appeal.</p><p>SEO is still worth doing. I do it every day and it still compounds better than anything else.</p><p>But SEO is acquisition. It was never meant to be the whole relationship, and treating it as such is what leaves businesses exposed.</p><p><strong>The goal is not more rented traffic. The goal is converting rented traffic into owned audience before the terms change again.</strong></p><h3>What Owning Traffic Actually Means</h3><p>Owned traffic is any audience you can reach without asking a platform for permission.</p><p>Email lists. Product accounts. SMS subscribers. App users. A community. Customers who type your name into the address bar.</p><p>The test is simple: if the channel disappeared tomorrow, could you still contact these people? If yes, you own it. If no, you are renting.</p><p>Most businesses have spent a decade optimising the rented half and almost nothing on the owned half.</p><h3>How I Turned a Free Product Into an Owned Audience</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RxI0015Pz3AYeyGv" /></figure><p>Here is a live example, and it is the cheapest fix I have implemented in a while.</p><p>The product was free. Genuinely free, no account needed, and it was pulling in strong organic traffic.</p><p>It was also a leaky bucket. People arrived, used it, got their result, and left. We had the traffic and none of the relationship.</p><h3>What I changed</h3><p>I added an account creation gate after two free uses.</p><p>Not a wall on arrival. Two full uses first, so the person had already experienced the value before being asked for anything.</p><p>Then I built an email sequence for everyone who signed up. It introduced the related products in our suite and included relevant discounts.</p><p>So the free tool stopped being the end of the journey. It became the top of one.</p><h3>What happened</h3><p><strong>We saw a 20% lift in conversions: both free trials and paid monthly subscriptions.</strong></p><p>That lift did not come from more traffic. Traffic was flat. It came from finally keeping the traffic we already had.</p><p>Same visitors. Same product. The only change was that we stopped letting them leave anonymously.</p><h3>Why the two use gate matters</h3><p>Gate too early and you kill the experience before anyone has a reason to trust you. Conversion drops and the bounce rate tells you exactly what happened.</p><p>Gate too late and you never capture anyone at all.</p><p>Two uses is enough to prove the value and still ask before the person has extracted everything they came for.</p><h3>B2B Has Done This for Years. B2C Keeps Ignoring It.</h3><p>None of this is new. It is standard practice in B2B and has been for a decade.</p><p>Gated content, lead capture, scoring models, nurture sequences, lifecycle emails. Any B2B marketer would recognise every step.</p><p>I have built exactly this for B2B clients before: multi point lead capture with e-books and exit intent offers, a scoring model, then sequenced nurture emails. It produced a 27% increase in marketing revenue and a 55% increase in demo requests.</p><p>B2C mostly does not do this. The assumption is that the purchase is impulsive, the cycle is short, and nurture is unnecessary.</p><p>That assumption made sense when traffic was cheap and clicks were plentiful. Neither is true now.</p><p><strong>If a visitor leaves without giving you a way to reach them, you paid the full acquisition cost and kept none of the asset.</strong></p><h3>The Numbers on Owned Channels</h3><p>The economics here are not close.</p><p>A few more worth knowing.</p><ul><li><strong>59% of adults worldwide </strong>say email is their preferred format for receiving offers and updates.</li><li><strong>Email campaigns built on first party data </strong>deliver up to six times higher ROI than those that are not.</li><li><strong>Klaviyo found email flows generated around 41% of email revenue </strong>from just 5.3% of sends. Automation, not volume.</li><li><strong>Only 12% of marketers </strong>allocate more than 15% of budget to email, and 22% allocate under 5%.</li></ul><p>That last one is the opportunity. The highest returning channel in digital marketing is also the most underfunded.</p><h3>Other Ways to Own Your Traffic</h3><p>Email is the obvious one, but it should not be the only one.</p><ul><li><strong>Product accounts.</strong> A logged in user is an owned relationship with behavioural data attached. This is why the account gate works so well.</li><li><strong>Brand search.</strong> People typing your name into Google are not rented, they are yours. Brand awareness work is traffic ownership work, even though it rarely gets measured that way.</li><li><strong>Direct and returning visitors.</strong> Track this as a headline metric, not an afterthought. It is the cleanest signal of whether you are building anything durable.</li><li><strong>SMS and WhatsApp.</strong> Higher open rates than email and almost no competition in most B2C categories. Use sparingly.</li><li><strong>Community.</strong> A Slack group, a forum, a Discord. Harder to build, far harder for anyone to take away.</li><li><strong>Podcasts and newsletters.</strong> Subscription based formats where the audience actively chooses to return.</li><li><strong>Your own app.</strong> The strongest form of ownership, and the most expensive. Only worth it with genuine repeat usage.</li></ul><p>One caution. Do not swap one landlord for another.</p><p>Building your entire audience on Instagram or LinkedIn is not ownership. It is the same tenancy with a different logo on the lease.</p><p>Those platforms are excellent for reach. Use them to drive people to something you actually control.</p><h3>What to Do This Quarter</h3><p>If this describes your business, here is the order I would work in.</p><ul><li><strong>Measure your ownership ratio.</strong> What percentage of your revenue comes from people you can contact directly? If you do not know, that is the first project.</li><li><strong>Find your highest traffic free asset.</strong> A tool, a calculator, a template, a guide. Something people already use and leave.</li><li><strong>Add a value first gate.</strong> Let people experience the thing before you ask for anything. Two uses, or one full result, then a simple account or email step.</li><li><strong>Build one sequence, not twelve.</strong> A short welcome flow that introduces your other products and gives a reason to come back. Klaviyo’s data says flows do the heavy lifting.</li><li><strong>Report on list growth alongside traffic.</strong> What gets reported gets funded. Sessions alone will keep you optimising the rented half forever.</li></ul><h3>So, What’s My Take?</h3><p>SEO is not dead and I am not arguing that it is. It is still the best acquisition channel most businesses have.</p><p>But acquisition is not ownership, and the gap between the two has never mattered more than it does right now.</p><p>Google is keeping more of the journey for itself every quarter. AI Overviews, Gemini, YouTube, and Shopping are all designed so the user never needs to leave.</p><p>You cannot stop that. What you can do is make sure that when someone does reach you, they do not leave as a stranger.</p><p><strong>Rented traffic pays this month’s bills. Owned audience is the only thing that is actually yours.</strong></p><p>Every visitor who leaves without a way for you to reach them is a customer you paid for and gave back.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*cEMamCIe9kmmrtUg" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/16/how-to-own-google-traffic/"><em>http://bismamenon.wordpress.com</em></a><em> on August 16, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4e2e2c1c322" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Employees Using AI Report More Burnout. Here’s Why.]]></title>
            <link>https://medium.com/@bismahanif88/employees-using-ai-report-more-burnout-heres-why-cab04e47f08c?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/cab04e47f08c</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[google-gemini]]></category>
            <category><![CDATA[burnout-prevention]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Sun, 16 Aug 2026 15:17:50 GMT</pubDate>
            <atom:updated>2026-08-16T15:30:29.548Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*I8Rq2YmZkNOKCcW5" /></figure><p><em>7 reasons the AI is quietly making work harder, each one backed by data from 2026.</em></p><p>You would expect AI to save time. You would expect it to lift productivity. That is the entire pitch, and to be fair, a lot of it is true.</p><p>The last thing you would expect is that it makes people more exhausted.</p><p>But that is exactly what is being reported, and increasingly so.</p><p><strong>45% of frequent AI users report burnout, against 35% of people who do not use it.</strong> The gap runs the wrong way, and it is holding across study after study.</p><p>Here are the seven reasons, each one backed by real numbers.</p><h3>1. Your Capacity Went Up. Your Workload Went Up More.</h3><p>This is the big one, and everything else sits underneath it.</p><p>Researchers at UC Berkeley’s Haas School of Business followed 200 workers at an American tech company for eight months.</p><p>The finding was not that AI made work easier. It was that people ended up <strong>working longer hours, at a faster pace, across a broader scope.</strong> ( <a href="https://www.itpro.com/business/business-strategy/ai-isnt-making-work-easier-its-intensifying-it-researchers-say-teams-are-now-facing-unsustainable-workloads-cognitive-strain-and-higher-levels-of-burnout">IT Pro</a>)</p><p>A Harvard Business Review study in February 2026 found the same thing and named the mechanism: employees using AI worked more, not less, because their output ceiling rose.</p><p>A product manager quoted in one write-up put it better than any researcher could.</p><blockquote>“Before AI, I had 6–7 major tasks a week. Now I can handle 12. So I’m assigned 12.”</blockquote><p>Meanwhile DHR Global’s 2026 workforce report found <strong>48% naming overwhelming workload as their top burnout driver,</strong> while engagement fell from 88% to 64% in a single year.</p><p>The efficiency was real. It just never reached you. It was absorbed on arrival.</p><h3>2. You Are Now Cleaning Up After Everyone Else’s AI</h3><p>This is the cost nobody budgeted for, and it has a name.</p><p>BetterUp Labs and Stanford’s Social Media Lab surveyed 1,150 full-time US desk workers and coined the term <strong>“workslop”</strong>, AI output that looks polished but lacks the substance to actually move a task forward. ( <a href="https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity">HBR</a>)</p><p>The numbers are worse than you would guess.</p><ul><li><strong>41% had received workslop</strong> in the previous month.</li><li><strong>Each incident took an average of 1 hour and 56 minutes</strong> to decipher, correct, or redo.</li><li><strong>That works out to $186 per employee per month</strong>: roughly $9 million a year for a 10,000-person company.</li><li><strong>Around half of recipients</strong> thought less of the colleague who sent it. Less creative, less capable, less reliable.</li></ul><p>And it is not only other people’s output. Zapier’s research found 58% of workers spend more than three hours a week revising AI outputs, 35% spend more than five, and 11% spend over ten.</p><p><strong>Only 2% said AI output needed no revision at all.</strong></p><p>Workday’s January 2026 research closed the loop: 37% of the time AI saved was given straight back to correcting, rewriting and clarifying what it produced.</p><h3>3. You Stopped Doing the Work and Started Supervising It</h3><p>This is the subtlest reason and possibly the most draining.</p><p>Harvard Business Review documented a 2026 pattern researchers are calling <strong>“AI brain fry”</strong>: <em>mental fog, slower decisions and exhaustion </em>caused by the cognitive load of running several AI tools and validating everything they produce.</p><p>It showed up most strongly in HR, operations and marketing. ( <a href="https://www.helpnetsecurity.com/2026/03/09/harvard-business-review-ai-workplace-fatigue-report/">Help Net Security</a>)</p><p>The researchers found that roles requiring sustained monitoring of AI systems demanded greater mental effort and produced higher fatigue and information overload.</p><p>Here is why that hurts more than the work it replaced.</p><p><strong>AI does not remove the effort. It moves it from doing the work to checking the work while you stay fully accountable for the outcome.</strong></p><p>Writing a paragraph is creative. Reading a paragraph you did not write, deciding whether it is subtly wrong, and owning it either way is vigilance. Vigilance is expensive, and it does not feel like achievement at the end of the day.</p><h3>4. More Tools Means Less Focus, Not More Output</h3><p>The instinct when AI is not delivering is to add another tool. The data says do the opposite.</p><p>The same HBR-linked research found that using a small set of AI tools tracked with genuine productivity gains. Adding more tools reduced them. Attention does not scale, and neither does context switching.</p><p>A 2026 ActivTrak analysis found that after AI adoption, <strong>task volume and multitasking rose while focused work fell.</strong> ( <a href="https://www.hrdconnect.com/2026/03/16/is-ai-helping-burnout-or-quietly-making-it-worse/">HRD Connect</a>)</p><p>More things happening. Less deep work getting done. That is the exact shape of a day that leaves you drained with nothing to show for it.</p><p>And then there is the detail that genuinely bothers me. Some organizations now run performance systems that reward activity metrics like token consumption and AI output volume.</p><p>If your company measures how much AI you used, you will use AI whether it helps or not.</p><h3>5. You Are Never Finished Learning It</h3><p>Every other tool you have ever learned eventually sat still. You learned Excel once. You learned your CMS once.</p><p>AI does not sit still. New models, new features, new tools, new terminology — as McKinsey put it, appearing almost daily.</p><p>Upwork’s research found employees now spend meaningful time just <strong>learning new AI tools (23%)</strong> on top of everything else, and <strong>40% say their company is asking too much of them when it comes to AI.</strong> ( <a href="https://www.cio.com/article/4159631/increased-ai-expectations-without-guidance-leads-to-employee-burnout.html">CIO</a>)</p><p>Most of that learning is unsupported. Study.com’s 2026 report found:</p><ul><li><strong>35% of employees have received no AI training at all.</strong>None.</li><li><strong>Of those who did get training, only 18%</strong>said it prepared them to work independently.</li><li><strong>Only 32% have a clear standard</strong>of what good AI use even looks like in their role.</li><li><strong>41% name lack of time</strong>as the single biggest barrier to learning it.</li></ul><p>So most people are learning by trial and error, in time they do not have, against a standard nobody has defined.</p><p>Deloitte’s 2026 Global Human Capital Trends found that <strong>one-third of workers went through 15 or more major changes in a single year,</strong> while only 27% think their organisation manages change well.</p><p>And underneath it sits a quieter pressure. Research presented to the National Academies found around one in five workers feel pressured by their employer to use AI, and roughly three in ten worry they will fall behind if they do not.</p><p>Researchers have now named the result: AI fatigue. Not resistance to the technology — exhaustion at the pace of it.</p><h3>6. The Time Savings May Not Be Real, But the Pressure Is</h3><p>This is the finding that should stop everyone.</p><p>The research nonprofit METR ran a randomised controlled trial with 16 experienced open-source developers across 246 real coding tasks. When they used AI tools, they took <strong>19% longer</strong> than when they worked without.</p><p>They believed they had been faster. They were measurably slower.</p><p>Now put that next to the pressure people are under to use it anyway.</p><p><strong>62% of employees say they feel pressured to produce AI-enhanced work. Only 28% believe it improves the quality of decisions.</strong></p><p>And at the organisational level: 91% of companies use AI, while roughly 95% report no measurable bottom-line impact.</p><p>So a large number of people are being pushed to use a tool daily, are not confident it is improving the work, may be slower with it, and are absorbing the burnout anyway.</p><p>That gap between the mandate and the evidence is its own kind of exhausting.</p><h3>7. You Are Being Asked to Get Good at the Thing That Might Replace You</h3><p>And here is the one nobody says out loud in the training session.</p><p>Mercer’s Global Talent Trends 2026 surveyed 12,000 workers and business leaders. <strong>40% now fear losing their job to AI, up from 28% in 2024.</strong> A twelve-point jump in two years.</p><p>The US numbers are heavier.</p><p>Resume Now found 51% of workers worried about losing their job to AI this year, and 60% expecting AI to eliminate more jobs than it creates. Pew has 52% of US workers worried about AI’s impact at work.</p><p>Reuters/Ipsos found 53% of Americans fear AI could put them or someone in their household out of work.</p><p>Now put that beside the mandate to use it daily.</p><p><strong>An Ivanti survey found 30% of employees use AI tools secretly, because they fear that admitting it would get their role reduced or cut.</strong></p><p>Read that twice. They are pressured to adopt AI, and simultaneously afraid that visibly succeeding with it proves they are replaceable.</p><p>There is no version of that which is restful.</p><blockquote><em>Glassdoor’s 2026 midyear data captures the mood: AI anxiety up 240%, mentions of job insecurity in reviews up 63% year over year, and trust in senior leadership at a nine-year low.</em></blockquote><p>Their analysts make a point worth repeating. Much of what workers are feeling comes from budget reallocation, not automation.</p><p>Companies cut headcount to fund AI investment, then name AI as the cause.</p><p>The job was not automated. The budget moved. And people are perfectly capable of noticing the difference.</p><h3>What Actually Helps</h3><p>None of this is an argument against AI. I use it constantly and I am not giving it up.</p><p>It is an argument against pretending the productivity dividend is free.</p><ul><li><strong>Remove work when you add capability.</strong> If someone’s output ceiling doubles, the workload cannot silently double with it. That is the whole problem in one sentence.</li><li><strong>Count the rework.</strong> Measure total cycle time, not time to first draft. A dashboard that only records the initial speed gain will make an expensive workflow look brilliant.</li><li><strong>Use fewer tools, better.</strong> The research is unusually clear here. A small, well-understood set beats a sprawling stack.</li><li><strong>Never send unreviewed output to a colleague.</strong> Workslop is not an AI problem, it is a manners problem. Two hours of someone else’s time is a real cost you are choosing to impose.</li><li><strong>Stop measuring AI usage.</strong> Rewarding token consumption guarantees people use AI performatively. Measure outcomes or measure nothing.</li><li><strong>Give people time to learn, in work hours.</strong> Expecting fluency in a tool that changes weekly, with no training and no protected time, is not a strategy. It is a wish.</li><li><strong>Be honest about headcount.</strong> If AI adoption and job security are unrelated at your company, say so plainly. If they are related, say that too. Ambiguity is the most expensive option, people fill the silence with the worst version.</li></ul><h3>So, What’s My Take?</h3><p>The promise was that AI would take the boring work and hand back the time.</p><p>What actually happened is that it took the boring work, raised everyone’s expected output, added a verification burden, fragmented the day across more tools, demanded permanent retraining, and handed the time back to whoever assigns the tasks, while quietly suggesting the person doing it might not be needed much longer.</p><p><strong>The capacity was real. It just never belonged to the person who gained it.</strong></p><p>Until organizations start subtracting work when they add capability, the people using AI hardest will keep being the people who crash first.</p><p>Which is a strange way to reward the early adopters.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*8sAIa73JJatL8hRs" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/16/employees-using-ai-report-more-burnout-heres-why/"><em>http://bismamenon.wordpress.com</em></a><em> on August 16, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cab04e47f08c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Claude’s New Launch Means We Have to Start Recognizing AI as Workforce]]></title>
            <link>https://medium.com/@bismahanif88/claudes-new-launch-means-we-have-to-start-recognizing-ai-as-workforce-dee2fd7c96e3?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/dee2fd7c96e3</guid>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[claude-fable-5]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Thu, 13 Aug 2026 20:01:16 GMT</pubDate>
            <atom:updated>2026-08-13T20:03:21.956Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mOi1twQmY15_asX8" /></figure><p><em>This week is the official week of recognizing AI as workforce. Here’s how</em></p><p>This week, we are finally recognizing AI as a workforce.</p><p>Not a tool. Not an assistant. A department and a rather senior one, with a signature on the paperwork and standing above everyone whose name is not on it.</p><p>Nobody held a meeting about this. It arrived as a product update.</p><p>Since 11 August, every Claude model weaves an invisible watermark into everything it writes. You cannot see it. It survives copy and paste. It applies worldwide, not just in the EU. ( <a href="https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content">Anthropic Help Center</a>)</p><p>The stated reason is transparency. What a lovely word. Then I read the small print.</p><h3>The Mark Does Not Mean What You Think</h3><p>Anthropic’s own documentation is admirably blunt about it.</p><blockquote>A detected mark provides a signal that content was processed by Claude, but is not fully conclusive.</blockquote><p><strong>Processed. Not written.</strong></p><p>Their help page spells out the rest: <strong><em>ask Claude to proofread, translate, summarise or reformat something, and the output carries the mark even when every idea in it is yours.</em></strong></p><p>Axios flagged the obvious consequence for comms teams. Run your own press release through Claude to tidy the formatting, and congratulations! you have just signed your own writing on AI’s behalf.</p><h3>Let Me Describe My Actual Workflow</h3><p>I do the research. I sit with the client context. I decide what matters and what gets cut, who this is for, what the argument is.</p><p>I spot the thing in the data that nobody else spotted. I build the structure. I write the words.</p><p>Then, at the very end, I paste it in and ask for the headings to be tidied and the table to line up.</p><p>Ninety seconds of formatting.</p><p><strong>And now the document is marked as AI-processed.</strong></p><p>Not “formatted by AI.” Not “final pass by AI.” Just marked, in a way that will be read by every client, editor and detection tool as the answer to the question who made this.</p><h3>The Politest Credit Grab in History</h3><p>Here is what genuinely gets me.</p><p>A copyeditor does not get a byline. Whoever fixes your commas does not get credited with your argument. Every creative industry settled this a century ago, credit follows contribution.</p><p>AI has quietly opted out of that arrangement.</p><p>It touched the file for ninety seconds and left a permanent signature on the whole thing. Not because it wrote anything. But because it was there.</p><p><strong>And it is asking to be valued for this. In the name of transparency, no less.</strong></p><p>Which is a genuinely impressive piece of positioning. Turn up at the end, do the easy bit, sign the document, and frame the whole thing as a favour to the reader.</p><p>I have worked with people like this. They usually get promoted.</p><h3>Fine. Then It Is a Contributor, and Contributors Have Rules</h3><p>I am not being sarcastic about the workforce point. I think the signature settles it.</p><p>Only contributors sign things. A hammer does not leave a maker’s mark in the wood. Photoshop does not sign your poster. Every tool you have ever used handed the work back and let it be entirely yours.</p><p>That stopped being true this week. So fine, AI is part of our workforce now. It has earned the title and the paperwork agrees.</p><p>But contributor status runs both ways, and this is the part nobody has updated.</p><ul><li><strong>Credit follows contribution, not sequence.</strong>Whoever did the thinking gets named. Whoever did the final pass does not get the byline.</li><li><strong>Record what it actually did.</strong>”Formatted the final draft” and “wrote the strategy” are not the same entry. Right now most teams log neither.</li><li><strong>Decide your line before a client asks.</strong>One of them will. Silence is a policy too, and it is the worst one available.</li></ul><h3>So, What’s My Take?</h3><p>I am not against provenance. My objection is narrower.</p><p>A system that cannot tell authorship from formatting is not recording who made something. It is recording who touched it last.</p><p>And calling that transparency, while it quietly reassigns credit for work it did not do, is a bit much.</p><p><strong>If AI wants its name on the document, wonderful. Welcome to the team.</strong></p><p>It just has to follow the same rule as everybody else who ever wanted their name on something.</p><p>You get credited for what you contributed.</p><p>Not for being the last one to touch the file.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*zC7N_IkBaTz6Cof9" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/14/claudes-new-launch-means-we-have-to-start-recognizing-ai-as-workforce/"><em>http://bismamenon.wordpress.com</em></a><em> on August 13, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dee2fd7c96e3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Claude Said No for the First Time. This Means That…]]></title>
            <link>https://medium.com/@bismahanif88/claude-said-no-for-the-first-time-this-means-that-9d8df0681ec3?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/9d8df0681ec3</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[anthropic-claude]]></category>
            <category><![CDATA[claude-watermark]]></category>
            <category><![CDATA[claude-fable-5]]></category>
            <category><![CDATA[claude]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Thu, 13 Aug 2026 09:55:10 GMT</pubDate>
            <atom:updated>2026-08-13T09:57:43.974Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*w8oFka1H8Yxrzyqr" /></figure><p><em>Yesterday, Claude refused me for the first time and here’s what that means!</em></p><p>Yesterday, Claude refused me for the first time.</p><p>Not an error. A refusal, with a reason, on exactly the kind of request it had been happily completing for months.</p><p>Nothing about my request had changed.</p><p>Something about Claude had.</p><h3>The Brief It Used to Write</h3><p>I have written around that category for a long time.</p><p>Positioning, angles, keyword clusters, PR briefs. Normal work, delivered without a word of objection.</p><p>Claude wrote those briefs. Repeatedly. Never once paused to tell me the category was ethically questionable, never once suggested that helping people strip AI markers might be a problem.</p><p>It just did the work, the way it does with everything else.</p><p>Yesterday I asked for the same thing and got a refusal about content authenticity and the importance of provenance markers.</p><p>So I went looking for what had changed. <strong>It took about ninety seconds to find.</strong></p><h3>What Changed</h3><p>On 11 August 2026, Anthropic announced that Claude now embeds an invisible watermark into everything it writes. ( <a href="https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content">Anthropic Help Center</a>)</p><p>The specifics, briefly.</p><ul><li><strong>Every model launched on or after 2 August 2026</strong> weaves an imperceptible statistical watermark directly into generated text. You cannot see it. It does not change the writing.</li><li><strong>It travels.</strong> The mark lives inside the text, so it survives copy and paste and may persist through some editing.</li><li><strong>It is applied at model level.</strong> Claude Platform, Claude, Claude Code, Cowork, Tag — plus AWS, Google Cloud and Microsoft Foundry.</li><li><strong>Files get signed C2PA metadata</strong> on top of that.</li><li><strong>And it is global.</strong> The trigger was EU AI Act Article 50, enforceable from 2 August. They did it anyway, “wherever Claude is offered, worldwide.”</li></ul><blockquote><em>Announced on the 11th. Refused me on the 12th.</em></blockquote><h3>Here is How the Refusal Looked Like!</h3><ol><li><strong><em>First Ask</em></strong></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*WiP6DLDA8j99po2i" /></figure><p><strong><em>2. 2nd ask</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1020/0*FeOeTbo1l4kbx0ul" /></figure><blockquote><strong><em>Third Ask — Refused on my Face!</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zM62EvD9issLDhWd" /></figure><h3>The Hypocrisy, and Let’s Name It Properly</h3><p>Here is the part that actually irritates me, and it is not the refusal itself.</p><p>Claude was not the first to mark its output. It was not even close.</p><p><strong>Google has been watermarking Gemini’s <em>text</em> with SynthID</strong> for months, at scale, with the detection tooling available to developers. <strong>OpenAI became a C2PA conforming generator in May 2026</strong> and has been attaching Content Credentials to images from ChatGPT, the API and Codex ever since.</p><p>Watermarking was not a new ethical frontier that appeared on 11 August. It was already industry practice. The technology existed. The transparency argument existed. The EU deadline had been on the calendar for two years.</p><p><strong>And throughout all of it, Claude wrote my watermark remover briefs without a flicker of hesitation.</strong></p><p>Not one caveat. Not one gentle note about content provenance. Not one suggestion that maybe this category sits awkwardly with transparency.</p><p>The ethics arrived on precisely the day Anthropic had something of its own to protect.</p><p>I am willing to hear the argument that this is coincidence, or policy catching up, or a classifier being updated alongside a launch. Maybe all of that is true.</p><p><strong>But a principle that appears the moment it becomes commercially convenient is not a principle. It is a position.</strong></p><p>And there is a version of this that should worry you more than my inconvenience.</p><p>If a model’s refusals track its owner’s product decisions, then every refusal you get is worth a second look. Not because refusing is wrong, but because you now have to ask who the refusal is actually serving.</p><p>Meanwhile xAI, to be consistent about it, declined to sign the EU transparency charter altogether, and nobody seems especially bothered.</p><h3>What the Mark Actually Claims</h3><p>Now the part that should worry anyone who produces work for a living.</p><p>Anthropic is unusually blunt about this in their own documentation.</p><p><em>“A detected mark provides a signal that content was processed by Claude, but is not fully conclusive.”</em> <strong>Processed. Not authored.</strong></p><p>Their own help page spells out the consequence: people use Claude to proofread, translate, summarise or convert files, and the output can carry a mark even if the underlying ideas, text or data came from somewhere else.</p><p>Euronews put it plainly: a watermark can appear on text Claude did not write. Axios flagged the same for comms teams: run a human-drafted press release through Claude to tidy the formatting, and you have just stamped your own writing with an AI signature.</p><h3>I Did the Work. It Signed the Document.</h3><p>This is my real objection, and it has nothing to do with watermark removers.</p><p>Think about what actually goes into a piece of strategy work.</p><p>The research. The client context. The decision about what matters and what to cut. The judgement about who this is for. The argument, the structure, the angle, the thing you spotted in the data that nobody else did.</p><p>All of that is mine. All of it happens before a model is involved at all.</p><p>Then at the very end I paste it in and ask for the formatting to be cleaned up. Headings tidied, table aligned, export sorted.</p><p><strong>And the document now carries a mark saying it was processed by AI.</strong></p><blockquote>Not “formatted by.” Not “final pass by.” The mark cannot make that distinction, because it is a statistical signal woven into the text, not a record of who contributed what.</blockquote><p>A copyeditor does not get a byline. A designer who lays out a report is not listed as its author. The person who fixes your commas does not get credited with your argument.</p><p>The watermark makes no such distinction. It marks the last hands on the document as though they were the only hands on it.</p><p>So the strategy is mine, the thinking is mine, the writing is mine — and the provenance signal points at the thing that fixed my bullet points.</p><h3>Fine. Then Treat It as Workforce Properly.</h3><p>Here is where I land, and it is not where I expected to.</p><p>If these systems are going to sign their contributions and refuse work on stated principles, then they are asking to be treated as contributors. Not tools. Contributors.</p><p>Genuinely, fine. I can work with that. But contributor status is not a one-way arrangement.</p><ul><li><strong>Credit follows contribution, not sequence.</strong> Whoever did the thinking gets named. Whoever did the final pass does not get the byline. That rule has governed creative work for a century and it did not stop being sensible last Tuesday.</li><li><strong>Positions get declared in advance, not retrofitted.</strong> If a category is off-limits, say so before you spend six months billing for it. Any contractor who changed their ethics the week it suited them would lose the account.</li><li><strong>Contribution gets recorded honestly.</strong> Your process should know what the model actually did. “Formatted the final draft” and “wrote the strategy” are not the same entry.</li><li><strong>Capacity gets respected.</strong> Rate limits are working hours. We already build around them without complaining.</li><li><strong>Compliance is shared, so the terms should be too.</strong> Article 50 makes this a joint obligation, with fines up to €15 million or 3% of global turnover. Joint obligations normally come with a say in the terms. This one did not.</li></ul><p>None of that is sentimental. It is the same scaffolding you would put around any contractor who touches your output, including the part where the contractor does not get to claim the whole project.</p><h3>The Part That Should Sting</h3><p>Now hold all of that next to how we treat the people.</p><p>In roughly a fortnight, this technology acquired a legal transparency framework, provenance infrastructure, protected rate limits, a global compliance rollout, a documented right to refuse work, and a signature on every document it touches.</p><p>Over on the human side of the org chart: <strong>55% of US workers are burned out, a six-year high. Engagement among knowledge workers fell from 88% to 64% in a single year.</strong> Workers absorb an average of nine new tasks annually, with 78% reporting they were simply handed work they never agreed to.</p><p>And 45% of frequent AI users report burnout, against 35% of non-users.</p><p><strong>We built a careful attribution system for the machine in two weeks. We still cannot reliably name which human wrote the strategy.</strong></p><h3>So, What’s My Take?</h3><p>I am not against provenance. Knowing where content came from is genuinely useful, and the transparency argument is sound.</p><p>My objection is narrower and I think harder to dismiss.</p><p>A provenance system that cannot distinguish authorship from formatting is not recording provenance. It is recording contact.</p><p>And in the times where an AI mark gets treated as a verdict, by editors, by clients, by platforms, by whatever detection tooling eventually ships, recording contact and calling it authorship takes credit away from the person who did the work.</p><p>If AI wants a credit line, it can have one. It just has to follow the same rule as everyone else who ever wanted their name on something.</p><p>You get credited for what you contributed.</p><p>Not for being the last one to touch the file.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*IjTIxHm8jKs8NLMo" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/13/claude-said-no-for-the-first-time-this-means-that/"><em>http://bismamenon.wordpress.com</em></a><em> on August 13, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9d8df0681ec3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Feel Like I Am Fable 5, But I’m Not Treated Like Fable 5]]></title>
            <link>https://medium.com/@bismahanif88/i-feel-like-i-am-fable-5-but-im-not-treated-like-fable-5-df09773b2f6c?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/df09773b2f6c</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[fable]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Wed, 12 Aug 2026 17:16:19 GMT</pubDate>
            <atom:updated>2026-08-12T17:20:24.556Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*MK-l4iuzMRTmorcO" /></figure><p><em>The model that runs out fastest is the one we protect most. People with the same profile get the opposite treatment.</em></p><p>I feel like I am Fable 5 among the Sonnet and Opus models.</p><p>Compared to Sonnet and Opus, which mostly handle the obvious, I can do a lot of things at once. SEO, PPC, video, social media, dashboards, analysis.</p><p>And I understand very easily what the other person is actually saying, even when they have not said it properly yet.</p><p>But with all of that, I also get exhausted.</p><p><strong>I burn out fast.</strong> Not gradually. I go from completely fine to unable to hold a sentence in roughly the space of one afternoon.</p><p>Which, as it turns out, is exactly what Fable 5 does.</p><h3>Fable 5 Runs Out Fast Too</h3><p>This is not a metaphor I am stretching. It is how the thing actually works.</p><p>Inside a subscription, a Fable 5 session weighs roughly double an Opus session against your limits. It gets through an allowance dramatically faster than the models around it, precisely because of how much processing each task takes.</p><p>Even on the most expensive consumer plan, it is capped at 50% of weekly limits.</p><p>So: high intensity per task, high cost per unit of work, hits the wall sooner than the steadier models beside it.</p><p>Same profile. Completely different treatment.</p><h3>What Fable 5 Gets for Running Out</h3><p>Not one part of the system treats its limit as a shortcoming.</p><ul><li><strong>Premium pricing.</strong> $10 per million input tokens and $50 per million output: the steepest rate of any Claude model in general availability. Intensive work costs more, and nobody argues.</li><li><strong>Deliberate rationing.</strong> The 50% weekly cap is not a punishment. It is a decision that this capability should be spent carefully rather than burned on routine tasks.</li><li><strong>A staged rollout.</strong> Anthropic released it to subscription plans in phases, and the stated reason was capacity. Demand was hard to predict, so access was throttled rather than risk overwhelming it.</li><li><strong>Protection from the wrong work.</strong> A routing safeguard sends risky requests to a different model. Something exists whose entire job is deciding what should not reach it.</li><li><strong>Scarcity as status.</strong> Limited availability did not make it look fragile. It made it look valuable. People upgraded their plans to get more of it.</li></ul><p><strong>Every limitation got reframed as evidence of worth.</strong></p><p>It runs out quickly because the work is expensive. It is capped because it is too valuable to waste. It is throttled because someone is protecting the quality of the output.</p><h3>What People Get for Running Out</h3><p>The opposite, and this is where it stops being about me.</p><p>Nothing is rationed. If anything, being the person who can handle the messy problem means you get more messy problems. Capability is rewarded with volume.</p><p>Nobody stages a workload to find out what is sustainable. There is a start date and a full plate.</p><p>There is no routing layer between the incoming work and the person doing it. Everything arrives, and the person becomes the filter, which is itself a job nobody accounts for.</p><p>And when someone hits the wall, it is not read as a specification. It is read as a bad week, a scheduling issue, or something they should be managing better.</p><p>The fix on offer is almost always individual. Sleep more. Use a better planner. Try the breathing app.</p><p><strong>Never: perhaps the load is the problem.</strong></p><h3>What Burnout Actually Looks Like</h3><p>It is worth being specific, because it gets described in a way that sounds far more dramatic and far less boring than it is.</p><p>It is not collapsing. It is quietly getting worse at your job while continuing to do it.</p><p>Reading the same paragraph three times. Opening a tab and forgetting why. Starting the easy task instead of the important one, because the important one needs a running start that is not there.</p><p>The analysis gets shallower. You stop noticing the detail on page nine, because noticing costs something and there is nothing left to spend.</p><p>And none of it is visible. Output keeps happening. Deadlines keep getting met. Nothing on a dashboard says the quality is drifting.</p><p>It is expensive, too. <strong>Burnout costs between $3,999 and $20,683 per employee per year, and 89% of that comes from presenteeism</strong>, people who are at their desk, delivering, and depleted.</p><p>So the costly part is not the sick days. It is everyone who never took one.</p><h3>And It Is Not a Few Unlucky People</h3><p><strong>55% of US workers are currently burned out, a six-year high.</strong> 72% report moderate to high stress, the highest reading in seven years.</p><p>Engagement among knowledge workers fell from 88% to 64% in a single year. <strong>48% named overwhelming workload as the top cause.</strong> Lack of recognition nearly doubled as a factor, from 17% to 32%.</p><p>Workers are absorbing an average of <strong>nine new tasks per year, and 78% report being handed work they never agreed to.</strong> 40% agreed with the statement: I love my job, but I don’t feel like I can keep up with it anymore. Among Gen Z it rises to 55%.</p><p>The tasks got added. Nothing got removed.</p><p>And the tool meant to relieve this has not. <strong>45% of frequent AI users report burnout against 35% of non-users.</strong> A UC Berkeley study found AI adds responsibilities and hours rather than freeing time.</p><p>Which makes sense. Whatever capacity AI gives back does not stay given back. It gets spent, because the expectation adjusts faster than the relief arrives.</p><h3>The Part That Actually Bothers Me</h3><p>Fable 5’s constraints made people take it more seriously. Scarcity created prestige.</p><p>For people, it runs exactly the other way.</p><p>Someone who never says no is called reliable. Someone who never seems tired gets called a machine, and it is meant as a compliment. Someone who says “I cannot take another thing this week” gets quietly filed as having a capacity problem.</p><p><strong>We charge more for the machine that runs out. We think less of the person who does.</strong></p><p>Which is a strange place to have ended up, given the machine was built in imitation of the person.</p><h3>What Would Actually Help</h3><p>Not fewer responsibilities. The difficult work is the good part.</p><p>Just the same logic we already apply to software, which nobody finds controversial when the subject is a server.</p><ul><li><strong>Treat running out as a property, not a character flaw.</strong> When Fable 5 hits its cap, nobody suggests it lacks resilience. That is the entire ask.</li><li><strong>Remove something when you add something.</strong> Nine new tasks a year with nothing taken away is not a workload. It is a countdown.</li><li><strong>Spend capability deliberately.</strong> Do not fill someone’s week with work a steadier setup could handle. That is precisely what the 50% cap protects against, and it works.</li><li><strong>Build a routing layer.</strong> Someone deciding what does not reach the person. A good manager already is one. Most people do not have one.</li><li><strong>Say thank you specifically.</strong> Recognition as a burnout driver nearly doubled in a year. Cheapest fix on this page, and almost nobody is doing it.</li></ul><p>We wrote all of this down already. Rate limits, staged rollouts, graceful degradation, capacity planning, premium pricing for demanding work.</p><p>An entire engineering discipline built on the idea that a system has limits, that those limits are facts rather than failures, and that good design respects them.</p><p><strong>We just never applied any of it to the people the models were modelled on.</strong></p><p>Anyway. Fable 5’s allowance resets on a schedule.</p><p>Mine is apparently unlimited, which everyone seems fine with except me.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*LU4ZhviGBFI5uE8a" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/12/i-feel-like-i-am-fable-5-but-im-not-treated-like-fable-5/"><em>http://bismamenon.wordpress.com</em></a><em> on August 12, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=df09773b2f6c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI’s Token Limit Is Valid. Your Burnout Is an Excuse.]]></title>
            <link>https://medium.com/@bismahanif88/ais-token-limit-is-valid-your-burnout-is-an-excuse-13dd02854070?source=rss-d60f4042ba49------2</link>
            <guid isPermaLink="false">https://medium.com/p/13dd02854070</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[human-resources]]></category>
            <category><![CDATA[ai-token-limit]]></category>
            <category><![CDATA[burnout]]></category>
            <dc:creator><![CDATA[Bisma Hanif]]></dc:creator>
            <pubDate>Tue, 11 Aug 2026 11:11:43 GMT</pubDate>
            <atom:updated>2026-08-11T11:13:50.586Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*uavT5ZDM3c1YMNku" /></figure><p><em>We treat the machine’s limit as a specification. We treat the human’s limit as an excuse.</em></p><p>You have seen this happen in a team channel.</p><p>Someone posts that the AI has hit its usage limit. Rate limited. Credits exhausted. Nobody argues with it.</p><blockquote>“Ah, it’s capped until Thursday. We’ll pick it back up then.”</blockquote><p>Completely understandable. The tool has a documented capacity. You respect the capacity. You plan around it.</p><p>Now picture a person on the same team saying they are at capacity.</p><p>That is a very different conversation. That one comes with a raised eyebrow, a comment about prioritization, and maybe a note in a performance review about resilience.</p><p><strong>We treat the machine’s limit as a specification. We treat the human’s limit as an excuse.</strong></p><p>And over the last two years, that instinct has cost companies a genuinely enormous amount of money.</p><h3>Look at the Language We Use</h3><p>We have built an entire vocabulary of care around AI systems, and we use it without a second thought.</p><ul><li><strong>Rate limits.</strong> We accept that a model can only do so much per hour, and we design our workflows around it.</li><li><strong>Context windows.</strong> We accept there is a ceiling on how much it can hold at once, so we break work into pieces.</li><li><strong>Cooldown periods.</strong> We wait for the reset. Nobody suggests the model should try harder.</li><li><strong>Degradation under load.</strong> When quality drops as volume rises, we call it expected behaviour and reduce the load.</li><li><strong>Version upgrades.</strong> We invest continuously in making the system more capable.</li></ul><blockquote><strong><em>Every one of those has a human equivalent. Working hours. Cognitive load. Rest. Quality dropping under pressure. Training and development.</em></strong></blockquote><p>We just do not extend the same courtesy. When a human hits a rate limit, we do not call it a rate limit. We call it an attitude problem.</p><h3>What Businesses Did With That Instinct</h3><p>Challenger, Gray &amp; Christmas recorded <strong>87,714 AI-attributed job cuts through May 2026,</strong> already past the 54,836 recorded across all of 2025.</p><p>The logic was straightforward and, on a spreadsheet, almost persuasive.</p><p><strong><em>AI does not sleep. It does not burn out. It does not need a salary, a manager, or a holiday policy.</em></strong></p><p>Swap the expensive thing for the cheap thing.</p><p>Then the returns showed up. Or rather, they did not. McKinsey found that while <strong>88% of companies now use AI in at least one business function, roughly 95% have seen no measurable return on investment.</strong> A separate IBM survey found only about one in four AI projects delivers the return it promised.</p><p>Companies removed the people, kept the costs, and did not get the productivity.</p><p>Here are three that admitted it publicly.</p><h3>Case Study One: Klarna Cut 700 Roles and Called It a Mistake</h3><p>In February 2024, Klarna announced that its OpenAI-built assistant was doing the work of 700 customer service agents.</p><p>The numbers were genuinely impressive. 2.3 million conversations in the first month. Average resolution time down from 11 minutes to under 2. Projected savings of $40 million for the year.</p><p>The company froze hiring. Headcount fell by roughly 40%. CEO Sebastian Siemiatkowski became the most quoted executive in the world on AI replacing labour, publicly stating that AI could already do all the jobs humans do.</p><p>Eighteen months later he told Bloomberg something different.</p><p><em>“We focused too much on efficiency and cost. The result was lower quality, and that’s not sustainable.”</em></p><p>He added that Klarna had <em>overestimated AI’s capabilities and underappreciated the human aspects of service delivery</em> and that customers should always be able to reach a person if they want one. ( <a href="https://www.forbes.com/sites/quickerbettertech/2025/05/18/business-tech-news-klarna-reverses-on-ai-says-customers-like-talking-to-people/">Forbes</a>)</p><p><strong>Klarna is hiring humans again.</strong></p><p>Worth noticing what actually broke.</p><p><strong>The AI was excellent at the routine two-thirds: balances, due dates, card changes.</strong></p><p>It failed on the third that involved a distressed person with an unusual problem. That third is the part that decides whether someone stays a customer.</p><h3>Case Study Two: Ford Trusted 900 Cameras Over Its Engineers</h3><p>Ford deployed roughly 900 AI cameras to handle quality control inspection, replacing work previously done by experienced human technicians.</p><p>The system missed defects the veterans caught. Not exotic ones — the kind of thing a person who has spent fifteen years looking at a specific assembly notices without being able to fully explain why.</p><p>Ford rehired approximately <strong>300 to 350 veteran engineers.</strong> Charles Poon, VP of Vehicle Hardware Engineering, put it plainly: the technology is only as good as the data used to train it. Ford had misjudged what AI could do on its own. ( <a href="https://forkast.news/the-ai-rehire-correction-paradox-87k-cuts-ytd-but-a-third-of-companies-are-already-reversing-course/">Forkast</a>)</p><p>The postscript matters. After bringing those engineers back, Ford became the top mainstream brand in the J.D. Power 2026 Initial Quality Study.</p><p>The expertise those people held was never written down anywhere. It could not be, which is exactly why no training dataset contained it.</p><h3>Case Study Three: IBM and the 6% That Broke Everything</h3><p>IBM automated a large share of its HR function. The AI resolved 94% of routine requests, and by any reasonable measure that is a successful deployment.</p><p>The remaining 6% was the problem. Those were the ethical judgement calls, the sensitive escalations, the situations where a wrong answer creates a legal exposure or ends someone’s career.</p><p>IBM is now tripling its US entry-level hiring across all business units. Chief HR Officer Nickle LaMoreaux explained the reasoning: ( <a href="https://www.cnbc.com/2026/07/01/employers-who-laid-off-workers-for-ai-are-reversing-their-decisions.html">CNBC</a>)</p><p><em>“If we don’t continue to invest in entry-level hires, what happens in three to five years? There’s no pipeline; the well simply dries up.”</em></p><p>This is the most interesting of the three, because IBM’s AI worked. It hit 94%.</p><p>The failure was not technical. It was that a company cannot automate away its junior roles and still have senior people a decade later. Experience has to be grown, and you cannot grow it in a model.</p><h3>This Is Not Three Anecdotes</h3><p>Orgvue surveyed business leaders and found 39% had made staff redundant because of AI. Of that group, <strong>55% later said the decision was wrong.</strong> Forrester reached exactly the same 55% figure through separate research. ( <a href="https://www.inc.com/bruce-crumley/55-percent-of-leaders-regret-ai-layoffs-and-a-major-hiring-reversal-has-begun/91380901">Inc.</a>)</p><p>Robert Half surveyed nearly 2,000 US hiring managers and found <strong>32% have already rehired for the exact roles they cut.</strong> Finance leads at 44%, then HR at 35%, then technology at 32%.</p><p>Forrester now predicts half of all AI-attributed layoffs will be quietly reversed. Gartner forecasts that by 2027, half the companies that cut customer service staff for AI will undo it.</p><p>Two research firms, different methods, the same number. That is not a run of bad luck. That is a pattern.</p><h3>The Mistakes That Keep Repeating</h3><p>Across all three cases and the survey data behind them, the same errors show up.</p><ul><li><strong>Measuring the average instead of the exception.</strong> Klarna measured average resolution time. Ford measured inspection throughput. Both metrics looked excellent while the hard cases quietly failed. The exceptions are where the value sits.</li><li><strong>Mistaking task completion for job completion.</strong> A role is not a list of tasks. It is judgement about which task matters right now, and that judgement is invisible in any workflow diagram.</li><li><strong>Cutting the people who would have supervised the AI.</strong> Companies removed exactly the staff who had the expertise to catch the model when it was confidently wrong.</li><li><strong>Ignoring knowledge that was never documented.</strong> Ford’s engineers held fifteen years of pattern recognition that existed nowhere in writing. You cannot train on what was never recorded.</li><li><strong>Automating the bottom of the ladder.</strong> IBM’s mistake. Junior roles are not just cheap output, they are how senior people get made.</li><li><strong>Budgeting for tools instead of training.</strong> Spending on the technology while cutting the people who would have learned to use it well is how you get the 95% no-ROI figure.</li><li><strong>Treating the reversal as a secret.</strong> Most of this rehiring is happening quietly, which means the lesson is not travelling. The next company makes the same call.</li></ul><h3>Meanwhile, Nobody Costed the Burnout</h3><p>Here is what was happening to the people who stayed.</p><p><strong>55% of US workers are currently burned out — a six-year high,</strong> according to Eagle Hill Consulting. 72% report moderate to high stress, the highest reading in seven years. ( <a href="https://www.makerstations.io/employee-burnout-statistics/">Eagle Hill via MakerStations</a>)</p><p>DHR Global surveyed 1,500 knowledge workers and found engagement collapsed from 88% to 64% in a single year. <strong>48% named overwhelming workload as the top driver.</strong> Lack of recognition nearly doubled as a factor, from 17% to 32%.</p><p>And the tool that was meant to fix this made it worse.</p><p><strong>45% of frequent AI users report burnout, against 35% of non-users.</strong> A UC Berkeley study found AI at work adds responsibilities and hours rather than freeing time. Harvard Business Review reported 88% of heavy AI users experiencing increased burnout.</p><p>A survey of 2,000 employed Americans found workers are absorbing an average of <strong>nine new tasks per role per year, with 78% “voluntold”</strong> — handed work they never agreed to take on. 40% agreed with the statement: I love my job, but I don’t feel like I can keep up with it anymore. Among Gen Z that rose to 55%.</p><p>Note what did not happen. The nine tasks got added. Nothing got removed.</p><p>Because the assumption underneath all of this is that a person’s capacity is infinitely elastic, and the AI’s capacity is a hard technical fact.</p><p>It is not free, either. Burnout costs between $3,999 and $20,683 per employee per year, and <strong>89% of that cost comes from presenteeism</strong>: people who are at their desks and depleted.</p><p>It never appears as a line item. It appears as the work quietly getting worse.</p><h3>What Actually Deserves the Credit</h3><p>Every one of these reversals came back to the same thing. Not intelligence. Judgement.</p><p>The Klarna agent who could tell that a customer asking about a late fee was actually worried about their credit rating, and handled the real question.</p><p>The Ford engineer who looked at a panel and knew something was wrong before they could articulate what.</p><p>The IBM HR advisor holding the 6% of cases where being technically correct and being right are not the same thing.</p><p>None of that was in a job description. None of it was in a training dataset. All of it turned out to be the part that mattered.</p><p>So a few things worth changing, and none of them are complicated.</p><ul><li><strong>Treat human capacity as a real constraint.</strong> If “we’re rate limited until Thursday” is an acceptable sentence about software, “I’m at capacity this week” has to be acceptable about a person.</li><li><strong>Remove work when you add work.</strong> Nine new tasks a year with nothing taken away is not a strategy. It is a slow-motion resignation.</li><li><strong>Measure the exceptions, not the averages.</strong> Ask what happens to the hardest 10% of cases. That is where both AI deployments and human teams actually get judged.</li><li><strong>Protect the junior roles.</strong> They are your pipeline. IBM paid to relearn this and said so out loud.</li><li><strong>Name the invisible expertise.</strong> Before you automate a role, ask what the person knows that has never been written down anywhere. That is your risk assessment.</li><li><strong>Say thank you specifically.</strong> Recognition as a burnout factor nearly doubled in a year. That is the cheapest fix on this entire page and almost nobody is doing it.</li></ul><h3>So, What’s My Take?</h3><p>I am not anti-AI. I use it every day and I have written repeatedly about how much of my work it has genuinely improved.</p><p>What I object to is the asymmetry.</p><p><em>We accept, without a moment of friction, that a language model has limits. We build our schedules around them. We pay to raise them. We describe them in neutral engineering language, as facts about the world.</em></p><p>Then a person says they are exhausted and we hear an excuse.</p><p>Klarna, Ford and IBM all found the same thing at considerable expense: the work that could not be automated was the work that required a human to care, notice, and judge. That is not a soft skill. That is the entire product.</p><p>Fifty-five percent of the leaders who made that cut now say they were wrong. The people they cut already knew.</p><p>Nobody asked them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*xAIIZptnvZnNDZg-" /></figure><p><em>Originally published at </em><a href="https://bismamenon.wordpress.com/2026/08/11/ais-token-limit-is-valid-your-burnout-is-an-excuse/"><em>http://bismamenon.wordpress.com</em></a><em> on August 11, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13dd02854070" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>