We like to think the biggest risk of AI is that it sometimes gets things wrong.

But if you actually use it every day, you know that’s not the real problem.
The real problem is that AI is very good at being just convincing enough.
It gives you answers that feel clean, structured and usable.
Even when the input is incomplete, the output looks like something you can just… move on with.

And that’s where it gets uncomfortable.
Because this isn’t just an AI problem.
Our brains work the same way.
We fill in the gaps. We smooth things over. We turn fragments into stories that make sense, even when they shouldn’t.

In this session, I explore that pattern through a set of recognisable and personal cases, including one where a series of child interviews slowly shifted from vague and uncertain to detailed and convincing, without anyone deliberately making things up.
It’s not hard to see the same thing happening when we work with AI today.
We’ll look at what’s going on underneath: confirmation bias, cognitive offloading, suggestibility, and why we trust fluent answers more than accurate ones, especially when we’re under time pressure.

This is not an “AI is bad” talk. I use AI myself. A lot.
But once I started noticing this pattern, I also started changing how I use it, where I pause, what I double-check, and where I no longer trust the first good answer.

Because the real shift is not only that AI can think with us, it’s that it quietly changes the point where we stop thinking for ourselves.

And if that moment feels familiar…
yeah.
It’s a trap.