The Barnum Effect Comes for AI
A couple of weeks ago I noticed that I’m not excited, but rather annoyed by people talking about AI stuff on social media. I, like many other people, felt FOMO and was scanning socials so I wouldn’t miss any of the new products and features. But the topic has become so overhyped and marketing has become so noisy that finding something actually useful that lives up to expectations has become almost impossible.
Markets for smart products existed before the AI boom, and one of the key tasks for any marketing team would be to showcase the product in a way that makes users want it while also aligning expectations.
And maybe in 2026 it’s the new norm to overpromise, but 5 to 10 years ago such behavior was called fraud. The recent case where Apple lost in court because they hadn’t delivered Apple Intelligence as they had presented it is one of the clearest examples that it’s not over yet.
Diving deeper
So I tried to understand how we got ourselves into such a position, and it turns out the issue I’m talking about is more of a combination of unfair practices and psychological loopholes that existed before but were never able to work so well.
So what am I talking about? It’s not a secret that most startups are trying to become the new Apple or Google. And the current top AI model developers are startups. They have brought a new kind of powerful technology to the market, and their main goal is for others to adopt it as a new daily tool — the way we adopted iPhones and Google search. OpenAI’s and Anthropic’s showcases are intentionally simple: they show a general direction and then pass the rest to your imagination. That, in combination with vague promises, leads to a situation where you, the end user, think about dozens of ways to use these tools. And this is the point where each of us creates our own unique and personal perception of AI products.
Why does it work?
But why does it work at all? Why do these vague promises make us believe that their models are the best and that they are capable of changing our daily lives so much that we are afraid of missing even the smallest feature? The answer has two parts to it.
Part I
The first part is that no one actually knows what AI and AGI are. Even the freaking AGI bench — the test specifically designed to mark the achievement of AGI — had to release a second version, and then a third, because AGI tasks were solved by non-AGI models. Despite this, Sam Altman, the CEO of OpenAI, keeps telling us that AGI will be achieved next year, this time for sure. And Dario Amodei, the CEO of Anthropic, says that all the code will be written by AI in six months. They sound more and more like fortune-tellers than experts with insider knowledge.
All of them abuse the same psychological loophole called the Barnum effect, a psychological phenomenon where individuals believe generic, vague personality descriptions apply specifically to them, despite the descriptions being applicable to almost everyone.
The same happens with promises from AI companies. We got lost in the Barnum effect around the terms “AI” and “AGI” because everyone just relies on their imagination instead of a strict list of use cases and capabilities. So in some way we got this “magic” that we were promised; unfortunately, it turned out to be more of a charlatan’s trick than the magic from Harry Potter.
Part II
The second part of the issue is that no one knows how to catch LLM developers lying, since the performance of AI models for most end users is kind of an abstract thing and mostly based on vibes. And a lot of us have already been tricked the same way by Apple.
Most people know that old iPhones become worse right after you update iOS to the latest version. Battery worse, performance worse, go on and buy a new one.
And you’ve probably noticed that, in the same way, right before the new frontier model comes out, the existing most powerful model starts to work worse. Everyone goes on Twitter and shares that Claude was lobotomized. But, again, we can’t prove it; we just have this nagging feeling that something is wrong.
Another frequent technique is cherry-picking. Companies will show you the best result they can achieve over hundreds of attempts instead of showing you the average result. And because of that, our expectations regarding new models and AI features are totally messed up.
And we kinda have to tolerate such behavior from these AI giants, because despite the overpromises we still got a big boost in productivity. We actually got that new tool that we use every day and that changed our workflow.
It’s contagious
But what we cannot tolerate is the same behavior from smaller non-AI products or from early-stage startups. At some point, having smart features became a mandatory checkpoint on every product roadmap. They force somekindofaifeature into their product. And when it’s time to promote, they do somekindofmarketing. So what they do is repeat after these giant AI providers. Again, they showcase something simple and promise something vague about power, performance, and potential. And that low-quality, low-value content fills social media on a daily basis.
Fake user-generated content abuses FOMO
What’s even worse is that they post fake user-generated content and flood Twitter, LinkedIn, and Threads with it. And that is where we get ourselves in a position where we think that we are losing to AI — that everyone is using some new tool that we are not. We feel left behind and FOMO kicks in. FOMO makes us try new products, pay $20 subs for each new one, or spend more tokens because of tokenmaxxxing. FOMO replaced excitement.
How to resist manipulations
But at the end of the day, when we go to meet our friends and family, we start talking with them about how bad we feel being left behind, and it turns out they feel the same. And we think, well, that is weird — everyone on Twitter says that it’s over. Then we decide to go to meetups and conferences and, again, the answer is the same. We find a couple of people who tried OpenClaw or Hermes or some other tool, but after a few questions it turns out that no, they have not automated their lives, they don’t have 50 automated businesses, and they still work at their jobs. They are just fans of trying new things, and the process itself brings them joy.
The next day, we stop checking every new tool, skill, or workflow. We start to see these FOMO manipulations made to farm impressions, and we start thinking critically. We finally feel relieved and have space for excitement to come back into our lives.