The truth about AI experts

Being skeptical isn't the same as dismissing everything. The right question isn't whether someone speaks well — it's whether they've solved a problem like yours and can show you how.

Why there's so much recycled content

AI became hype. And hype attracts people who want quick attention. The result is a sea of generic content: "10 prompts that will change your life," "How to make money with AI."

The problem isn't the topic. It's that most of these people have never used AI to solve a real problem. They tested it for 10 minutes and wrote an article.

There's also a strange incentive at work: content about AI circulates for being about AI, not for fixing anything. The weekend produces lists; Monday morning doesn't.

What people who actually use it know

People who use AI for real know uncomfortable things: where the model is confidently wrong, which tasks it handles badly, and which ones aren't worth attempting. They also know that no tool rescues a badly defined process.

And they show their work: the before and after of the same request, with the prompt they used. If someone won't show the prompt, you're looking at an opinion — not a method.

Healthy skepticism applies to the experts too: ask for the example. Whoever has one shows it; whoever doesn't changes the subject.

3 signs of quality content

1. Has concrete examples: not "AI can help you." It's "I used AI to solve this specific problem, and here's the before and after."
2. Shows limitations: everyone talks about what works. People who actually use AI also talk about what DOESN'T work — and why.
3. Is reproducible: you can copy the method and apply it to your context. Doesn't depend on a specific tool.

3 signs of recycled content

1. Promises "easy and fast" results: if it were that easy, everyone would already be doing it.
2. Only talks about generic prompts: "50 prompts" lists copied from other lists. Zero context, zero method.
3. No name, no face: anonymous or overly generic content with no accountability for what it says.

The simplest rule

Ask yourself: "has this person actually used AI to solve a problem like mine?" If the answer is no, it's probably recycled content.

If you like criteria written down, the public search quality evaluator guidelines are worth a read: they show how professionals are trained to judge content against explicit criteria instead of a "feels good" impression.

In practice, reproduce the prompt in your own context before adopting it. If it doesn't reproduce, it's an anecdote — and an anecdote isn't a method. And be wary of numbers with no source: ChatGPT making things up shows how that looks.

Frequently asked questions

Is everyone talking about AI trying to trick me?

No. Healthy skepticism is a filter, not a verdict. The criterion is method: whoever solved something, showed how, and owns up to what doesn't work is usually legitimate — even when the take is unpopular.

How do I know whether an expert actually knows?

Ask for the concrete example: which problem, which prompt, what came out, and what went wrong. No before and after means well-told theory. With one, you can judge by the result.

Are AI studies and reports worth anything?

They are, when they state the method and the scope. If a figure appears with no source and no explanation of how it was measured, treat it as packaged opinion. Don't repeat a number you haven't seen at its origin.

I don't have time to check everything — where do I start?

Pick a few sources that show their method, follow them for a while, and test it on your own case before adopting it. One cheap check a week beats ten articles skimmed in passing.

The shortest path is learning by doing: From beginner to pro lays out the three levels, and 10 ways to use AI at work gives you the first examples.