AI Washing: How Companies Are Faking Intelligence to Trick You
I bought a “smart” home security system last year that was supposed to use AI to detect threats. What it actually did was send me notifications when the wind blew the bushes. When I called support to complain, they told me the AI was “learning” and would improve. It never did. I realized months later that there was no AI at all. They had a motion sensor and a marketing budget.
🚨 HIRING: Tech Talent
💰 $50–$120/hr | 🔥 Multiple RolesFrontend • Backend • Full Stack • Mobile • AI/ML • DevOps
👉 Apply Here
That’s AI washing. It’s everywhere now. You see it in productivity apps that just built if-then automation and called it intelligent. You see it in recruiting software that uses pattern matching and claims to understand cultural fit. You see it in customer service chatbots that are essentially fancy search functions pretending to think. Companies slap the word AI on a thing that does something vaguely computational and suddenly investors show up and customers believe they’re getting something futuristic.
The worst part is that AI washing isn’t even that hard to do. Most people don’t know the difference between a chatbot, machine learning, and an actual AI system. They just see the word AI and think about the future. Companies know this. They’re counting on it.
I’ve spent the last two years watching companies do this. I’ve tested their products. I’ve read their marketing. I’ve talked to their engineers off the record. What I’ve figured out is that AI washing follows patterns. Once you see them, you can’t unsee them. And once you know what to look for, you realize how much of what’s being sold to you isn’t intelligent at all. It’s just expensive automation wearing an AI mask.
The Thing They’re Actually Doing When They Say AI
AI washing starts with a simple premise: people are scared of missing out on AI, so they’ll buy anything labeled with it.
Your email system probably has a “smart” inbox. What that means is it’s using rules. If the email contains certain words or comes from certain senders, it goes somewhere. That’s pattern matching. That’s not intelligence. But call it AI and suddenly it sounds like your computer is thinking for you. Call it machine learning and it sounds even better.
The most common form of AI washing is taking something that already existed and adding a machine learning layer that barely changes anything. Your email spam filter used to use keyword blacklists. Now it uses machine learning to understand which emails are spam. The difference is real, but it’s not magical. It’s still mostly pattern matching. Still mostly good at obvious spam and bad at subtle spam. Still occasionally flags real emails as junk. But now when it fails, you don’t blame the filter. You assume the AI is still learning.
Another common one: automation with a friendly face. You’ve seen those productivity tools that promise to save you hours per week by automating your tasks. What they actually do is let you set up if-then rules. If I get an email from this person, create a task. If that task is marked done, send a notification. These are chains of basic programming. There’s nothing intelligent about them. But they’re AI-powered, supposedly. They learned your workflow. They understand your process. What they actually did was listen to you tell them your process and then execute it. That’s not learning. That’s following instructions.
The third type is the most insidious. It’s when they take one of the actually good language models (usually from OpenAI, Claude, or another existing company) and rebrand it as their own AI. They put a nice interface on top of it and call it proprietary technology. They charge you money for access to something that’s already available or that you could access cheaper somewhere else. They’re not faking the intelligence. They’re faking that they built it.
How To Know You’re Being Washed
There are tells. Actual AI companies can explain what their model does, what data it was trained on, and what its actual limitations are. They’ll tell you what it’s bad at, not just what it’s good at.
AI washing companies do something different. They use vague language. They talk about how the AI is “learning” and “improving” without explaining what that means. They tell you it uses “advanced algorithms” or “machine learning” without getting specific. When you ask what training data they used, they get quiet or give you corporate-speak. When you ask what it can’t do, they change the subject.
Look at their claims. If they say their AI will “revolutionize” something or “transform” an industry, be suspicious. Real AI improvements are usually specific and boring. They’ll say something like, “We achieved 87 percent accuracy on fraud detection, which is up from 79 percent.” That’s concrete. That’s testable. When companies make sweeping promises about transformation and revolution, they’re usually trying to distract from the fact that the actual improvement is marginal or nonexistent.
Another tell: they keep saying their AI is getting smarter. This is the “learning” lie. Most machine learning models don’t learn after deployment. They’re trained once on data, then they do their job. They don’t wake up smarter tomorrow. If a company tells you their AI is learning from your usage and getting better, ask them how. Ask them what data it’s using. Ask them if you can opt out of that training. Most will admit they’re not actually doing that. They just wanted you to think they were.
Pay attention to what they’re actually automating. If it’s something a rule-based system could do, it probably doesn’t need AI. If a recruiting software is just matching keywords on your resume to keywords in a job description, that’s keyword matching. That’s not AI. If a scheduling tool is just looking for blocks of free time and booking meetings, that’s basic logic. Not AI. If a writing tool is just running your text through Grammarly and calling it intelligent assistance, that’s not AI either.
The most honest companies will also tell you about failure cases. They’ll mention where their AI struggles. They won’t pretend it’s perfect. AI washing companies pretend problems don’t exist, or they blame the user for not understanding how to use the tool correctly.
Why Companies Do This (It’s Not Complicated)
The honest answer is money and market perception.
AI is expensive to actually build. Training a real language model costs millions. Building a model that’s actually better than existing ones costs more. So most companies don’t do it. Instead, they do what’s cheaper: they take existing technology, wrap it in a nice interface, and call it AI. Or they build a simple feature — some automation, some pattern matching — and market it as AI. Or they license someone else’s AI and claim it as their own.
The reason it works is that investors are terrified of missing AI. Every company needs an AI story now. So when a company says “our product uses AI,” the financial incentive is suddenly there. Valuations go up. Users believe they’re buying the future. The company doesn’t have to actually deliver on the promise because by the time the user figures out the AI doesn’t work very well, the company has already been acquired or gone public or pivoted to the next thing.
I talked to an engineer at a “smart workplace” company who told me that maybe three percent of what they marketed as AI was actually machine learning. The rest was automation, statistics, and basic logic. She said it was known internally as “the AI layer” as a joke because it was so thin. She left the company because she was tired of lying in sales calls.
Get Amit Tripathi’s stories in your inbox
Join Medium for free to get updates from this writer.
For companies, AI washing is risk-free. If customers believe the AI is learning and improving, then when it doesn’t work, the company can just say the AI needs more time or more data. The customer can’t prove that wrong. By the time they get frustrated and leave, another wave of new customers who believe the marketing have signed up.
The only reason any company would be honest about what their AI can and can’t do is if honesty was good for business. It usually isn’t. The incentive structure rewards exaggeration.
What Happens When You Fall For It
You lose money. That’s the obvious part.
But it goes deeper. When you buy a product that’s been AI-washed, you’re not just losing money on something that doesn’t work. You’re also burning credibility on AI as a tool. You buy a “smart” tool that’s dumb, and you start to think AI is dumb. You use an AI writing assistant that just changes your words around, and you think AI can’t actually help with writing. You try an AI recruiting tool that picks the wrong candidates, and you think AI is racist or useless.
What you’re actually learning is that this specific company’s product is poorly built or dishonest. But it feels like you’re learning something about AI as a category. So you skip the actual AI tools that might help because you’ve been burned by AI washing.
It also warps what gets built. Real AI research is hard and expensive. So when AI washing is rewarded and gets funding and customers, the incentive shifts away from actual innovation toward clever marketing. Why spend millions building a real AI model when you can spend a thousand dollars building a tool that looks AI-powered and market it for a hundred dollars a month? The rational choice from a business perspective is the second option.
The result is a market full of fake AI products that work badly, real AI products that get lumped in with them, and customers who are cynical about the whole category. Nobody wins except the companies doing the washing.
How To Not Be An Idiot About This
The simplest rule: if a company can’t explain exactly what their AI does and how it does it, don’t buy from them.
Ask specific questions. What training data? What’s the error rate? What can it not do? If they get vague or defensive, that’s your answer. Real AI companies want you to understand the limitations because that’s how you use the tool correctly. AI washing companies want you to believe in magic.
Look at the actual feature. Strip away the marketing and describe what it does in simple terms. Can a rule-based system do this? Then it probably doesn’t need machine learning. Is it just pattern matching? Then calling it AI is probably a lie.
Check what people actually say about the product. Not reviews, because those are usually written by people who don’t know what they’re looking at. Check the Reddit communities or forums where power users hang out. They’ll tell you if the AI actually works or if it’s just automation with a good UI.
Pay attention to the comparison claims. If a company says their AI is better than competitors, ask them what they measured and how. If they can’t answer, it’s marketing. If they can answer, they’ll give you a specific metric. Take that seriously.
Most importantly, stay skeptical of things that sound magical. AI is useful. It’s powerful. But it’s not magic. If something promises to revolutionize your life with AI, it’s probably just doing something boring with a better interface. And that’s fine. Boring useful things are valuable. You don’t need the AI lie. The boring truth is usually enough.
You’re Going To Fall For This Anyway
I want to end by being honest about something: even knowing all of this, you’re going to get fooled.
I know what AI washing looks like, and I still bought that security system. I still subscribed to tools that promised AI help and delivered automation. I still occasionally see something labeled AI and want to believe in it before I think about whether it makes sense. The marketing is designed by people who are good at their jobs, and knowing the trick doesn’t make you immune to well-executed tricks.
What matters more than never being fooled is noticing when you have been fooled and getting out. If you buy something and it doesn’t actually work the way the marketing promised, don’t assume you’re using it wrong. Don’t assume the AI is learning. Don’t wait around hoping it improves. Cancel it. Get your money back if you can. Move on.
The companies doing this are counting on inertia. They’re counting on you keeping a subscription to something that doesn’t work because you paid for three months and might as well see if it improves. That’s how they win. That’s how they stay in business despite the fact that their product is oversold.
Be the person who cancels. Be the person who asks hard questions. Be the person who admits when something didn’t work instead of pretending it did. That’s not a lot, but it’s enough to make marketing less effective and honesty more profitable.
Thank you for being a part of the community
Before you go:
👉 Be sure to clap and follow the writer ️👏️️
👉 CodeToDeploy Tech Community is live on Discord — Join now!
Disclosure: This post includes affiliate and partnership links.
