<?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[Content Resources - Medium]]></title>
        <description><![CDATA[Practical articles on SEO, link building, digital PR, and AI search visibility. Case studies, field notes, and lessons from real campaigns. - Medium]]></description>
        <link>https://medium.com/content-resources?source=rss----cc9aedceab6f---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Content Resources - Medium</title>
            <link>https://medium.com/content-resources?source=rss----cc9aedceab6f---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 11:19:58 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/content-resources" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Most Expensive Links I Can Buy You Are Not Links]]></title>
            <link>https://medium.com/content-resources/the-most-expensive-links-i-can-buy-you-are-not-links-53a2fded9256?source=rss----cc9aedceab6f---4</link>
            <guid isPermaLink="false">https://medium.com/p/53a2fded9256</guid>
            <category><![CDATA[seo]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[ai-search]]></category>
            <category><![CDATA[digital-pr]]></category>
            <category><![CDATA[link-building]]></category>
            <dc:creator><![CDATA[Boris Dzhingarov]]></dc:creator>
            <pubDate>Tue, 08 Sep 2026 04:08:28 GMT</pubDate>
            <atom:updated>2026-09-08T04:08:30.025Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cDq1HCtzkmziQqE-y0x1ew.png" /></figure><p><em>We priced 16,625 publishers. The ones at the top of the market will take your money and hand you a nofollow.</em></p><p>I went into this expecting to confirm something obvious: that you pay more for a stronger site, and that the strongest sites give you the best links.</p><p>Half of that held up. The other half turned out to be the opposite of what I believed.</p><h3>What we actually counted</h3><p>ESBO Ltd has been building a publisher database since 2015, and it is the largest asset the company owns. In September I exported every site in it with an Ahrefs Domain Rating of 30 or higher and at least 2,000 monthly organic visits. That left 16,625 sites across 53 languages and 33 categories, with the price each one quoted us for a sponsored article.</p><p>Both filters are blunt. A Domain Rating can be manufactured and a readership cannot, which is a separate problem I wrote about in Forbes as <a href="https://www.forbes.com/councils/forbesagencycouncil/2026/09/02/how-to-tell-a-real-publication-from-a-content-farm-before-you-pitch/">how to tell a real publication from a content farm before you pitch</a>. What follows describes sites that sell placements. It does not describe sites that deserve your money.</p><p>Every number below has a sample size behind it, and none of it comes from a survey asking people what they think a link should cost.</p><h3>Authority sets the price, not the niche</h3><p>The median publisher wants $570 for a sponsored article. Half the market sits between $250 and $1,290. One site in ten comes in under $100 and one in nine asks $2,500 or more.</p><p>What moves that number is authority. The median is $355 at DR 30 to 39 and $1,840 at DR 80 and above, a 5.2x spread. Traffic behaves the same way. Sites with 2,000 to 5,000 monthly visits ask a median $390. The 425 sites with over a million visits ask $2,030.</p><p>The niche premium everyone talks about is smaller than the folklore. Among publishers that accept gambling content, the median surcharge over their standard rate is 24%, and only 16% charge double or more. I had previously put that premium at two to five times, which the data says is the top of the range and not the norm. Correcting your own numbers in public is character building, apparently.</p><h3>Then the part I did not expect</h3><p>One publisher in ten sells nofollow links only. Fine, that is a rounding error.</p><p>Except it is not evenly spread. At DR 80 and above, 39% of publishers sell nofollow only. Among sites with more than a million monthly visits, 36% do.</p><p>And they charge more for it. Nofollow-only publishers quote a median $1,840 against $510 for everyone else. That is 3.6 times the price for a link Google has been explicitly told to ignore.</p><p>Disclosure runs the same direction. Across the sample, 27% of publishers say they will attach a sponsored label. At DR 30 to 39 that is 15%. At DR 80 and above it is 57%. The publishers that label charge a median $1,300, and the ones that do not charge $430.</p><p>So the top of the market costs the most, gives you the least link equity, and is the most likely to mark your article as an ad. Every instinct I built in this industry over a decade says that should be a scam.</p><h3>It is not a scam. It is a different product.</h3><p>Large publishers do not sell link equity. They sell an audience and their name, and they price it the way they price advertising, because that is what it is. The nofollow attribute is not them cheating you. It is them telling you honestly what you bought.</p><p>For most of my career that would have made the placement worthless to me. It does not now, and the reason is the same thing this whole publication keeps circling.</p><p>When Ahrefs studied AI Overview visibility across 75,000 brands, branded web mentions correlated at 0.664 and backlinks at 0.218. Muck Rack, reading more than 25 million links cited by ChatGPT, Claude and Gemini, found earned media accounted for 84% of citations and paid or advertorial content for 0.3%. I made the fuller argument for that shift in Entrepreneur under <a href="https://www.entrepreneur.com/business-news/stop-counting-backlinks-start-counting-brand-mentions">Stop Counting Backlinks. Start Counting Brand Mentions</a>.</p><p>Put those next to each other and the expensive nofollow placement stops looking like a con. You are buying a sentence in a real publication that describes your company, in front of readers who might buy from you, in a source these systems read. The link was never the only asset. It was just the one we could measure, so it became the one we sold.</p><p>The uncomfortable half of that: the 0.3% figure suggests AI systems mostly skip what is marked as advertising, and the top of the market is exactly where the sponsored labels are. Digital PR, link building and brand mentions all point at the same conclusion here, which is that a placement people would have run anyway beats a placement you paid to have labelled.</p><h3>English is not one market</h3><p>The other finding I would act on if I were buying rather than selling.</p><p>The 6,164 English-language sites in the sample have a median price of $380, below the non-English median of $660. That number is close to useless on its own. Split English sites by where their readers actually live and the median is $1,330 for Australian audiences, $1,130 for Canadian, $1,040 for British, $452 for American and $100 for Indian.</p><p>Anyone quoting you a single English guest post price is averaging those together and hoping you do not ask.</p><p>Meanwhile Swedish sits at $1,440 and German at $1,230, both above almost anything you can buy in English, while Central and Eastern Europe runs a median of $490. If you sell in Warsaw or Bucharest, most English-language agencies have never looked at what you can get there for the price of one American placement.</p><h3>What I would do with a budget of $5,000</h3><p>Not nine placements at $570 each, which is what the median tempts you into.</p><p>Two or three at the top of a market where your buyers actually are, accepting that one of them will be nofollow and labelled, plus whatever earned coverage the same story can get with no money attached at all. Then track the mentions in the same document as the links, because the mention is the thing that keeps working after the campaign closes.</p><p>The full breakdown, with the price ladder by authority, all 31 language markets, the grey niche numbers and the link insertion data, is here: <a href="https://esbo.ltd/state-of-link-building-and-brand-mentions-2026/">State of Link Building and Brand Mentions 2026</a>. Every figure has its sample size next to it, and I have written down where the sample is biased, because it is our own network and it over-represents sites that are willing to sell you something.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=53a2fded9256" width="1" height="1" alt=""><hr><p><a href="https://medium.com/content-resources/the-most-expensive-links-i-can-buy-you-are-not-links-53a2fded9256">The Most Expensive Links I Can Buy You Are Not Links</a> was originally published in <a href="https://medium.com/content-resources">Content Resources</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Everyone Runs the AI Check Once. Almost Nobody Runs It Twice.]]></title>
            <link>https://medium.com/content-resources/everyone-runs-the-ai-check-once-almost-nobody-runs-it-twice-3e7faa01d393?source=rss----cc9aedceab6f---4</link>
            <guid isPermaLink="false">https://medium.com/p/3e7faa01d393</guid>
            <dc:creator><![CDATA[Boris Dzhingarov]]></dc:creator>
            <pubDate>Thu, 03 Sep 2026 06:34:24 GMT</pubDate>
            <atom:updated>2026-09-03T06:34:25.871Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A medium shot of a woman with her hair in a bun, wearing glasses and a tan knitted sweater, sits at an organized desk in a well-lit home office. She is looking intently at her open laptop, which displays a communication application, while her right hand uses a mouse. A second, larger monitor is positioned behind the laptop, showing a spreadsheet with data and a bar chart. Her left hand holds a pen and is poised over a checklist on a clipboard. The desk also features a coffee mug, a notebook with" src="https://cdn-images-1.medium.com/max/1024/1*fcLOuUP_tYSVglD8mhZkrg.jpeg" /></figure><p><em>The work is not difficult. It is boring, and boring is what loses.</em></p><p>After the last two pieces, the questions I got were about tactics. Does a Wikipedia page matter. Which assistant should I trust. How do I get started.</p><p>Nobody asked the one I would have asked, which is what happens in month four.</p><p>That is where this work dies. I have watched it happen with clients, and I have done it myself, so this is not a lecture.</p><h3>What month three looks like</h3><p>The first check is genuinely fun. You ask ChatGPT, Gemini, Copilot and Perplexity what they know about your company, somebody screenshots the worst answer, it goes in a Slack channel, and everyone has an opinion by lunchtime.</p><p>Month two you fix the obvious things. Old About page, three different versions of the company name, a directory profile written in 2021.</p><p>Month three, the answers have barely moved. The person running it has a launch on. The check becomes something you do before a big pitch rather than something you do on a date. Nobody cancels it. There is no meeting where the decision gets made. It just stops, and four months later somebody asks how the AI thing is going and nobody has an answer.</p><p>The reason I care about this is that a single reading is close to worthless.</p><h3>One screenshot tells you about one afternoon</h3><p>In the six month test I wrote about, changing the wording of the question could make the client appear or vanish. That alone should make anyone suspicious of a single result.</p><p>The ground underneath the answers moves too. Semrush tracked ChatGPT’s citation sources through 2025 and found that in the space of a few weeks in September, Wikipedia went from appearing in roughly 55% of prompt responses to under 20%, and Reddit collapsed from close to 60% to around 10%. Nobody’s brand did anything to cause that. The model changed its diet.</p><p>So a screenshot is trivia. Twelve screenshots taken the same way on the same dates is a trend line, and a trend line is the only thing that tells you whether your work did anything.</p><p>Which means the entire problem is a habit problem wearing a marketing costume. Everything I know about keeping this alive I learned from failing to keep other things alive, and I put the general version of that on my own blog a while back, on <a href="https://dzhingarov.com/building-new-habits-when-motivation-runs-out/">building new habits when motivation runs out</a>. The specific version, applied to this, is below.</p><h3>What we changed so it survives a bad month</h3><p>We made it smaller. Five prompts, four assistants, one document. It takes about twenty minutes and it produces no report and no deck. Every time we have tried to make it a proper deliverable, it has died within two cycles, because a deliverable needs someone in the mood to write it.</p><p>We attached it to something that already happens. It runs at the top of the monthly reporting call, not on its own calendar invite. Anything that requires its own slot competes with client work, and client work wins.</p><p>We took it off the person who cares most. The check is owned by someone junior with a checklist. Enthusiasm is a terrible scheduling mechanism. A checklist is a fine one.</p><p>We log it where everything else is logged. Every ESBO Ltd report already carries a mentions column next to the links, so the check output goes in the same document rather than a folder somebody has to remember exists. Digital PR, link building and brand mentions sit on one page, and the AI answers sit under them.</p><p>And we removed the requirement to feel good about it. Some months nothing moves. That is not a signal to stop, and it is also not a signal to try something clever. It is just what month five looks like.</p><h3>What still goes wrong</h3><p>I would like to say we solved this. We did not.</p><p>The check still slips when a big campaign lands. It is the first thing to go, every time, because nothing breaks when you skip it.</p><p>The worse failure was quieter. For the first few months we saved answers without saving the exact prompt that produced them. So we had a folder of screenshots we could not compare to each other, because nobody could remember whether June’s question had said “best providers of” or “leading providers of,” and as I said, that difference is enough to change the answer. Four months of data, mostly wasted.</p><p>Now the prompt text sits next to the answer, copied verbatim, with the date. It is the single most useful change we made and it is embarrassingly obvious in hindsight.</p><h3>Why the boring version is the whole game</h3><p>The tactics here are more settled than the discourse suggests. When Ahrefs looked at 75,000 brands, branded mentions across the web predicted AI visibility about three times better than backlinks did. Muck Rack, reading more than 25 million links cited by ChatGPT, Claude and Gemini, found 84% of citations came from earned media and 0.3% from paid and advertorial. You do not need a secret. You need other people writing about you, and you need to keep checking whether it landed.</p><p>Neither of those is difficult. Both are boring, and boring work does not get done on enthusiasm. It gets done because it is small, attached to something else, owned by a named person and written down in a place people already look.</p><p>If you take one thing from this: open a document, write your five prompts into it, run them the same way on the same day every month, and put the date at the top. In a year that document will be worth more to you than anything I or anyone else can tell you about the tactics.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3e7faa01d393" width="1" height="1" alt=""><hr><p><a href="https://medium.com/content-resources/everyone-runs-the-ai-check-once-almost-nobody-runs-it-twice-3e7faa01d393">Everyone Runs the AI Check Once. Almost Nobody Runs It Twice.</a> was originally published in <a href="https://medium.com/content-resources">Content Resources</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Spent Six Months Trying to Change What AI Says About a Client]]></title>
            <link>https://medium.com/content-resources/i-spent-six-months-trying-to-change-what-ai-says-about-a-client-d892bcec90ae?source=rss----cc9aedceab6f---4</link>
            <guid isPermaLink="false">https://medium.com/p/d892bcec90ae</guid>
            <dc:creator><![CDATA[Boris Dzhingarov]]></dc:creator>
            <pubDate>Sun, 30 Aug 2026 04:59:24 GMT</pubDate>
            <atom:updated>2026-08-30T04:59:25.382Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A man with glasses and a blue shirt sits at a desk in a home office, looking intently at a computer screen that displays a bar graph. He is using a mouse with his right hand. The desk is covered with notebooks, papers, a coffee mug, and a half-eaten pastry. The background shows bookshelves and a window." src="https://cdn-images-1.medium.com/max/1024/1*q-FOTlBWbKJ50RL1EVsJoA.jpeg" /></figure><p><em>Some of it worked. The parts that worked were not the parts I expected.</em></p><p>The last time I wrote about this, I described a habit we keep at ESBO Ltd: before a client walks into an important meeting, we ask ChatGPT, Gemini, Copilot and Perplexity what they know about the company, and we read the answers out loud. It is a cheap exercise and an uncomfortable one.</p><p>The question I got back, over and over, was the obvious one. Fine, but can you change it?</p><p>Six months ago I decided to stop guessing. One client, a B2B software company whose answers were a mess of outdated product descriptions and a competitor’s name, agreed to be the experiment. We tracked the answers monthly, screenshotted everything, and tried a list of things in parallel. Below is what happened, including the parts that make me look silly.</p><h3>The starting point</h3><p>Their profile had three separate problems, which I now think is typical.</p><p>The description was old. Two versions of the company were being blended into one, and the older one had better coverage online, so it kept winning. Second, the assistants disagreed with each other. Perplexity was closest to reality because it leans on recent pages and shows you its sources. ChatGPT was the most confidently wrong. Third, and worst, when I asked who the best providers in their category were, the client did not appear at all. Not lower down the list. Absent.</p><p>That third problem is the one worth caring about. Being described inaccurately is annoying. Not being named when a buyer asks who to consider is a lost deal you never hear about.</p><h3>What we changed</h3><p>We did four things, deliberately, so we could see which ones mattered.</p><p>We made every description of the company identical. Website, LinkedIn, every directory profile, every author bio. Same name, same spelling, same one-line description of what they do. This is unglamorous work that a junior person can do in a week, and it turned out to be the cheapest win available.</p><p>We deleted or rewrote the stale pages on their own site. Old product pages, an About page from before the pivot, launch announcements for a service they no longer sold. If your own site still tells the old story, you cannot blame a model for repeating it.</p><p>We went after coverage on other people’s sites. Interviews, expert commentary, a data study their team ran and gave away, a handful of podcast appearances. Some of that coverage linked to them and some did not, which used to bother me and now does not, for reasons I will get to.</p><p>And we tracked mentions the same way we track links. Every ESBO Ltd report now carries one extra column: where the company name appeared, with or without a hyperlink, and in what context. Brand mentions and backlinks sit next to each other, and clients read the mentions column first.</p><h3>What actually moved</h3><p>The description fixed itself first. Within about eight weeks, three of the four assistants were describing the current company rather than the 2022 version. The consistency work and the site cleanup are the obvious candidates, and I cannot separate them, because we did both at once. If I ran it again I would stagger them. Lesson noted, slightly too late to be useful.</p><p>The category question took longer and moved less. By month five the client started appearing in some phrasings of “who are the leading providers of X” but not others. Change the wording of the question slightly and they would vanish again. That instability is worth sitting with, because it means anyone showing you a screenshot of a perfect answer has also, somewhere, got a screenshot of a bad one.</p><p>What correlated best with the improvement was other people writing about them. That matches what the research says, which I went into in more detail in a piece for Entrepreneur about <a href="https://www.entrepreneur.com/business-news/stop-counting-backlinks-start-counting-brand-mentions">counting brand mentions instead of backlinks</a>. The short version: when Ahrefs looked at 75,000 brands, branded mentions across the web predicted AI visibility roughly three times better than backlinks did. Muck Rack, going through more than 25 million links cited by ChatGPT, Claude and Gemini, found that 84% of citations came from earned media and 0.3% from paid and advertorial content.</p><p>That 0.3% is the number I keep repeating to clients who want to buy their way out of this. You cannot. I have watched people try.</p><h3>What did not work</h3><p>Publishing more on their own blog did nothing I could measure. We shipped a lot of posts in those six months, and the assistants remained supremely uninterested. The Ahrefs research found the same thing at scale: how big your site is has almost no relationship with whether AI assistants mention you. The models care what other people say about you, not how often you say it about yourself.</p><p>Emailing publications to ask for a link on an existing mention worked about a third of the time. Worth doing, since it takes four minutes, but the mentions that stayed unlinked appear to have counted anyway.</p><p>And one thing I expected to matter did not seem to: their competitor with the strong Wikipedia page held on to the top of the category answers throughout. Six months of good work did not dislodge a decade of accumulated coverage. I do not think it was ever going to.</p><h3>Where this leaves link building</h3><p>I run ESBO Ltd, a digital PR and <a href="https://esbo.ltd/services/link-building/">link building agency</a>, so treat what follows with the skepticism it deserves, but the honest position is that the work matters more than the metric does.</p><p>Links still move rankings and still bring referral traffic. Nothing about that changed. What changed is what a placement is worth when the link does not appear. A year ago I would have logged that as a failed placement. Now I log it as a mention that will still be feeding an assistant’s answer long after the campaign closes. Same outreach, same relationships, same editors. Different scoreboard.</p><p>The practical effect on how we work is small but real. Digital PR and link building at ESBO Ltd now point at the same target: publications whose writing gets read and quoted, rather than whichever site has the tidiest metrics. We do not walk away from a placement because the editor’s policy strips links. And we ask for the link once, politely, then let it go.</p><h3>Six months later</h3><p>The client’s answers are better than they were and worse than they want. They show up in some category questions and not others. Their description is current. Their competitor still leads.</p><p>If someone offers you a fixed timeline and a guaranteed outcome for this, they are selling something I have not been able to buy. Part of what happens here is luck: which crawl caught which page, which source a model happens to weight this month, whether a writer who quoted you got syndicated. You can improve your odds without ever fully steering the result, and I have made peace with that better than my clients have.</p><p>What I can say after six months is that the boring end of the list did most of the work. Consistent naming, deleted stale pages, other people describing you in plain sentences. Nobody is going to build a software product around any of that, which is probably why the noisy end of the market keeps talking about everything else.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d892bcec90ae" width="1" height="1" alt=""><hr><p><a href="https://medium.com/content-resources/i-spent-six-months-trying-to-change-what-ai-says-about-a-client-d892bcec90ae">I Spent Six Months Trying to Change What AI Says About a Client</a> was originally published in <a href="https://medium.com/content-resources">Content Resources</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ask ChatGPT About Your Company Before Your Prospects Do]]></title>
            <link>https://medium.com/content-resources/ask-chatgpt-about-your-company-before-your-prospects-do-a3fa95781897?source=rss----cc9aedceab6f---4</link>
            <guid isPermaLink="false">https://medium.com/p/a3fa95781897</guid>
            <category><![CDATA[seo]]></category>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[branding]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Boris Dzhingarov]]></dc:creator>
            <pubDate>Fri, 14 Aug 2026 06:48:57 GMT</pubDate>
            <atom:updated>2026-08-14T06:48:58.412Z</atom:updated>
            <content:encoded><![CDATA[<p><em>I run this check before every client pitch. The answers still surprise me.</em></p><p>In early 2025 I was helping a founder get ready for a sales meeting. Big prospect, long shot, the usual nerves. Out of curiosity more than anything, I opened ChatGPT and typed his company name followed by a simple question: what does this company do, and are they any good?</p><p>The answer described a business that had stopped existing around 2022. Old product, old tagline, a pricing model they had abandoned. It also cited, with complete confidence, a directory listing that had three reviews on it. The founder read it over my shoulder and went quiet for a while.</p><p>We fixed what we could before the meeting, which was not much. But the habit stuck. Since then I have run the same check for every company we work with, across ChatGPT, Gemini, Copilot and Perplexity, before anything important (a pitch, a funding conversation, a hiring push). I keep the screenshots in a folder. Reading through that folder is one of the odder experiences my job offers.</p><h3>The check takes twenty minutes</h3><p>There is nothing clever about the method. I ask each assistant the same handful of questions. What is this company? Is it legitimate? Who are its competitors? Who are the best providers in its category? Then I ask about the founder by name, because buyers rarely research a company without also researching the person who emailed them.</p><p>Four assistants, five or six questions each, paste everything into a document. Twenty minutes, maybe thirty if the answers are messy enough to need a second pass.</p><p>What comes back falls into patterns I now recognize on sight.</p><p>The most common one is time lag. Assistants describe a composite of your past selves. A client of ours rebranded in 2023, and one answer used the old name in the first paragraph and the new name in the second, as if the two companies existed side by side. Another client killed a product line years ago and it still headlines their description, because the pages announcing the launch got far more coverage than the quiet removal ever did. Nobody writes articles about the things you stop doing.</p><p>Then there is the sourcing, which is where things get strange. A model will lean on a directory profile with three reviews as if it were an audited report. Perplexity at least shows its citations, so you can see where the smell is coming from. ChatGPT rounds everything into smooth, confident prose and leaves you guessing. I have seen answers built partly on a forum thread from years ago, started by someone with a grudge.</p><p>Mistaken identity comes up more than I expected. One of our clients shares a name with an unrelated firm on another continent, and some answers merge them into a single company that somehow does both things. If your name is not unusual, assume this is happening to you right now.</p><p>And then the category question, which is the one that stings. Ask who the best providers in a space are and watch a client who has done excellent work for a decade simply not appear, while a competitor with a well maintained Wikipedia page and a few podcast appearances anchors the whole answer. The first few times I watched a founder read that, I learned to schedule these sessions before lunch, not after.</p><h3>Nobody tells you it happened</h3><p>The reason the habit stuck is that buyers run this check too, and they never tell you. A procurement manager screening vendors asks an assistant which ones are credible. A marketing director sanity checks an agency before replying to a cold email. A journalist checks whether a quoted expert is real. There is no analytics report for any of it, no referral traffic, no bounce rate. The meeting just starts a few degrees warmer or colder than it should, and you never find out why.</p><p>I cannot prove how often this happens, and I distrust anyone who claims exact numbers. What I can say is that over the past year, prospects have started arriving at first calls with opinions that clearly came from a chat window rather than our website. A few of those opinions were about things we never published anywhere. You learn to hear the difference.</p><p>So pitch preparation at our agency now starts from the transcript. Before a founder walks into a meeting, we read out loud what the assistants say about them, all of it, the flattering parts and the nonsense. It is uncomfortable in a way that turns out to be useful. About half the time the founder interrupts with some version of: we have not done that since 2022. Right. And the person you are meeting on Thursday has no way of knowing that.</p><p>That changes the pitch itself. Instead of opening with the standard deck, you open by quietly correcting the specific wrong impression a machine has been distributing on your behalf. You never mention the machine. You just make sure the first ten minutes replace the stale story with the current one. It sounds small. It has changed the temperature of more meetings than any slide redesign we ever paid for.</p><h3>What moves the answers</h3><p>Can you change what the assistants say about you? Slowly, partially, and with less control than anyone selling a fix will admit. I have watched answers shift over months, and a few things kept showing up on the winning side.</p><p>Consistency, first. If your website, your LinkedIn page and your directory listings each describe a slightly different company, the model averages them, and the average is mush. Boring agreement between all of your own descriptions does more than one brilliant new tagline.</p><p>Being described by other people helps even more. A real writer explaining what you do in plain sentences, on a page crawlers read, outweighs your own boilerplate repeated across ten press releases. This sits close to my day job, so discount my enthusiasm accordingly, but the pattern holds no matter who does the work: third party descriptions are the raw material these systems trust most.</p><p>Deleting your own stale exhaust matters too. Old product pages, an About page from two pivots ago, launch announcements for things that no longer exist. Leave them up and you are feeding the composite self problem all by yourself.</p><p>And patience. Months, not weeks. Some answers move and some refuse to, and I have stopped pretending I always know why. Part of it is luck: which crawl caught which page, which source a given model happens to lean on that week. The honest version of this advice is that you can improve your odds without ever fully steering the outcome.</p><p>I still find the whole situation slightly ridiculous. A language model paraphrasing a directory listing from 2019 should not shape a sales meeting in 2026, yet I have watched it happen often enough to stop arguing about whether it should. Twenty minutes and a bruised ego. That is the full cost of finding out what the machines say about you while there is still time to talk back.</p><p><em>Boris Dzhingarov is the founder of </em><a href="https://esbo.ltd/"><em>ESBO Ltd,</em></a><em> a link building agency. He has been running AI reputation checks for client companies since early 2025.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a3fa95781897" width="1" height="1" alt=""><hr><p><a href="https://medium.com/content-resources/ask-chatgpt-about-your-company-before-your-prospects-do-a3fa95781897">Ask ChatGPT About Your Company Before Your Prospects Do</a> was originally published in <a href="https://medium.com/content-resources">Content Resources</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>