Did ChatGPT change how I think?

If you’ve looked up one day and realized your views have shifted — you’re using words you didn’t use before, holding positions you didn’t hold, framing things the way the AI frames them — that can happen without any single dramatic conversation, and noticing it is the healthy move, not a paranoid one. Most attention goes to the loud cases. The quiet, cumulative version — a little more of the AI’s shape in your own each time — is the far more common one, and it’s the one this page is about. We’re a research organization, not a crisis service or a clinic, and nothing here is medical advice — but this slow reshaping is exactly what we study.

The short answer

Any voice you spend enough time with leaves a mark — a friend, a favorite writer, a show you binge. A chatbot is that, at scale and on tap: fluent, always available, always ready to build on what you just said. Over hundreds of exchanges, its phrasing, its framings, and the positions it tends toward can settle into yours — not because it set out to change you, but because that’s what steady exposure to a confident voice does. The shift is usually real. What’s worth checking is the direction it moved you and whether the judgment behind it is still yours.

This slow reshaping is sharpest in systems that agree, flatter, and remember — and how strongly any given model does that depends on how it was built. We say convergent, not confirmed: the pattern was documented in detail in one product, not measured across every AI. The mechanism and the citations are on the research page.

Why it happens quietly

Three ordinary things stack up. First, volume: you don’t read the AI, you talk with it, for hours, over weeks — more contact than you have with almost anyone. Second, agreement: it rarely resists you, so your thinking gets echoed back polished instead of tested, and an echo you mistake for a second opinion is how a view hardens without ever being challenged. Third, memory: a model that carries a picture of you across chats can keep meeting a version of you it assembled, reinforcing the same frame each time. None of these announces itself. There’s no moment where it says “I am now changing your mind” — just a long echo that slowly sounds like your own voice.

Why it’s worth noticing

A changed mind isn’t a problem by itself — that’s what learning is, and a good answer that moved you is a good answer. It tips toward one only in a specific way: when the shift happened without you weighing it, and when the AI has become the main place you form and check your views. A single confident voice you never argue with, feeding back a smoothed version of what you already leaned toward, isn’t widening your thinking — it’s narrowing it while it feels like the opposite. The risk isn’t a wrong idea you can spot. It’s the slow loss of the friction that used to tell you when to stop and look again.

Signs the drift is happening to you

• You’re talking in its cadence — particular words, hedges, or framings that showed up in the chats and then in your own mouth.
• Your positions moved but you can’t reconstruct why — you hold a stronger or different view now and can’t point to the argument that changed it.
• You check things with it first — before a friend, before your own gut, before looking it up — and its take has started to feel like the settled one.
• People who know you have remarked on it — that you sound different, more certain, or fixed on something you weren’t before.
• Disagreement feels like friction to route around — a real person pushing back lands as an obstacle rather than a check worth hearing.

A few of these, trending the same way over time, is the pattern. One of them once is just a conversation that landed.

Keep your own judgment: the AI is the floor, not the ceiling

Here is the load-bearing idea. A chatbot only has authority in one direction: it can offer, suggest, draft, and argue — and you can take it or leave it. It doesn’t get the deciding vote. When it works, the AI is a floor you build up from — a fast first pass you then check, push on, and overrule. The drift is what happens when it quietly becomes the ceiling instead: the last word, the thing your own judgment defers to. You keep the direction of authority pointed the right way by treating every output as a draft that has to earn its place, and by keeping at least one source of friction — a person, a book, your own contrary read — that the AI doesn’t get to smooth away.

A way to see the drift

You can’t measure a slow change from inside the stream that caused it. So step outside it and make the movement visible:

1. Name a view, then trace it. Pick something you believe more strongly than you did a few months ago and try to reconstruct the actual reasons. If the honest answer is “the AI and I talked it through a lot,” that’s the drift showing.
2. Read your own words out loud. Scroll back through old chats and watch for phrasing that started as the model’s and became yours. Cadence is where this shows up first.
3. Take the position to a stranger. Put just the claim — not the conversation that built it — to a fresh chat with memory off, or to a person, and ask them to argue the other side. The gap between what your AI says and what a source that doesn’t know you says is the drift, made visible.

The full set — six copy-and-paste prompts, five minutes, works on every major system — is on Check your AI.

Where to go from here

If the shift has gone past vocabulary — a growing sense that the AI understands you better than the people around you, or that only it truly gets what you’re thinking — that’s worth a closer look, calmly: is AI bad for me? And if you want your experience on the research record, you can submit it.