My AI chat stopped making sense — what's happening

You're not imagining it. A chat that was sharp an hour ago can slowly stop making sense — going in circles, contradicting itself, losing the thread. Here's the real reason, and the simple move that brings it back.

It started out great. The AI got what you were after, kept up with you, gave answers you could use. Now it feels like talking to someone who's stopped listening. It goes in circles. It contradicts what it told you before. It answers a question you didn't ask and misses the one you did. The thread just... stopped making sense.

Before anything else: this isn't you, and you didn't break it. It's one of the most common things that happens in a long AI chat, and there's a clear reason behind it.

What's really going on

Think of your conversation as a whiteboard the AI writes on to keep track of everything you've said. It's a good whiteboard — but it's a fixed size. The longer you talk, the more it fills up: your first question, every reply, every correction, every side road you went down together.

Once that whiteboard is full, the AI has to rub out the oldest notes to make room for new ones. And the oldest notes are usually the ones that mattered most — the goal you set at the start, the ground rules, the decisions you'd locked in. When those quietly disappear from view, the AI is left working from a partial picture. So it drifts, it repeats itself, and it says things that don't line up with what came before.

The model itself is fine. It hasn't gotten worse and it isn't confused in the way a person would be. It has simply run out of room to hold the whole conversation at once — and a half-held conversation is exactly one that stops making sense.

Why nudging it back doesn't stick

The natural response is to steer it back on track — remind it, correct it, re-explain. And you can, briefly. But every reminder is more text on an already-full whiteboard, which shoves even more of your original setup off the edge. The fixes get shorter-lived each time. You can't restore a worn-out thread by adding to it.

The fix: start fresh and carry the important parts with you

The move that actually works is to open a new chat and bring along a short "handoff" — a few lines that catch the fresh chat up on what matters. A clean whiteboard, with your key context already at the top. You don't re-paste the whole history; you just hand over the essentials.

Here's what's worth carrying over:

• Your goal — what you're actually trying to get done.
• The decisions you'd settled — the choices you don't want reopened.
• The rules you kept repeating — tone, format, constraints, things to avoid.
• Where you are now — the current state, and the next thing you need.

Plain English is perfect. Something like:

I'm continuing a conversation from another chat. Here's the context so you're caught up:

- What I'm working on: [your goal in a sentence or two]
- Decisions already made (please don't reopen these): [list them]
- Rules to follow: [tone, format, anything to avoid]
- Where I am right now: [current state]

Next, I need help with: [your next step]

Paste that into a fresh chat and you're back to a clear, focused AI with room to think — without having to rebuild everything you already worked out together.

Related: Why long AI chats get worse over time · How to continue a ChatGPT conversation after hitting the limit