The Problem With AI Hype

I've been thinking about writing this one for a while because I was worried it would come off wrong. Like, I run a blog about AI tools. I genuinely get excited about this stuff. So if I write "the hype is out of control" does it seem hypocritical? Maybe. But I think it's actually more important to say it coming from someone who actually uses these tools every day, not someone who's just reacting from the outside.

Here's what triggered this. I was watching a founder demo on Twitter a few weeks ago, a new AI productivity tool, and they were making these claims that I, as someone who uses a dozen AI tools regularly, knew were either wildly overstated or just not how these systems work. And the comments were full of people losing their minds. "This changes everything." "I can't believe this is real." "My workflow will never be the same."

And I just felt tired. Not because the tool wasn't real or even that it wasn't useful. But because the framing was so disconnected from what the tool actually does in practice, and nobody in the thread seemed to be applying any skepticism at all. Everyone was just matching the founder's energy and amplifying it.

That's the hype machine. And it has real costs that I don't think we talk about enough.

What hype actually does to regular people

Okay so real talk: when something is hyped past its actual current capability, regular people try it, have a realistic experience (meaning: good but not magical), and feel like they did something wrong. Like they couldn't figure out how to open up the amazing thing everyone was describing.

I've talked to so many people who tried ChatGPT early on, had a pretty normal experience where it was sometimes helpful and sometimes confidently wrong, and walked away thinking "I must not be the target audience" or "I'm probably not smart enough to use it right." When actually the tool was just... not what the hype said it was. The hype set an expectation that made the reality look like failure. That's a bad outcome.

And then there's the flip side. When something is hyped and people have bad experiences, there's this backlash overcorrection where suddenly the tool is useless and AI is a fraud and it was all marketing. Neither the hype nor the backlash is accurate. Both make it harder to have a clear-eyed conversation about what these tools actually are and aren't.

I genuinely think hype culture has set AI adoption back in some ways. Because the people who would benefit most from practical AI tools, people with lots of repetitive work, people who need research help, people writing in a second language, are the ones who got burned by overpromising and wrote the whole thing off. While the people who were going to be early adopters regardless are already deep in the weeds. The hype reaches the wrong people in the wrong way.

Why the hype machine exists and who it serves

I'm not naive about this. AI companies need funding. Funding requires excitement. Excitement requires hype. The incentives are totally clear. A demo that shows a new feature doing something genuinely useful but modest doesn't go viral. A demo that shows something that looks kind of miraculous does. Even if the miracle required very specific conditions that won't apply to most users' actual situations.

The demo-to-reality gap is a known problem in software generally, but I think AI is especially bad for it because the outputs are so variable. A normal software demo shows you a feature working correctly. And it will work correctly the same way in your hands. An AI demo shows you a best-case output, and your output might be very different because the inputs are slightly different or the model is having a bad day or you asked it differently.

I've started applying a rule when I watch AI demos: I don't trust any single output. I want to see multiple attempts. I want to see the prompts, not just the results. I want to know what happened when it didn't work. Almost no demos give you this. Which tells you something about whose interests the demo is serving.

The influencer problem

Okay I have to be honest here because I'm in this ecosystem. There's a whole class of AI content creator, myself to some degree included, who benefits from excitement. More excitement means more clicks, more followers, more affiliate revenue if you have it. So there's this incentive to match the founder energy, amplify the demos, be the person who's always bringing good news about AI.

I try really hard not to do this. I know I don't always succeed. But I want to name it because I think it's a real distortion. The people creating content about AI tools are not a neutral sample. They're selected for enthusiasm. And enthusiasm is not the same as accuracy.

When I write about a tool I'm excited about, I try to also write about what it can't do and what frustrated me about it. Not to be balanced for the sake of balance, but because I genuinely think that's more useful information. If you're trying to decide whether to spend time learning a new tool, knowing where it falls apart is at least as important as knowing where it shines.

What I wish the conversation looked like

Honestly, I'd love to see more "I used this tool for real work for three weeks and here's what actually happened" content and less "watch this demo and imagine the possibilities" content. The former is harder to make. It requires patience and honest assessment. It doesn't go as viral. But it's what people actually need to make decisions.

I'd also love it if the AI company PR machine would stop saying things like "this will 10x your productivity" because it sets up new users to feel like failures when they see a 1.3x improvement, which is actually great! That's a meaningful improvement! But it doesn't feel great when you were promised 10x.

The tools are genuinely interesting and useful in their real-world form. They don't need the extra hype. The hype is for funding rounds and viral moments, not for helping people. And I think it's worth saying that clearly, even when you're someone who's genuinely enthusiastic about the underlying technology. Maybe especially then.