# ChatGPT Can't Or Won't Learn From Its Mistakes — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/e770a6bcef2e5f82

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-10-27 17:46 ingested 2026-10-28 02:35

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1xgzes6/chatgpt_cant_or_wont_learn_from_its_mistakes/)

## retained text



```
This happens to me often on ChatGPT.  I'm on the Plus plan:


&nbsp;


I'll present an issue to the AI then we discuss.  ChatGPT does not follow my instructions.  It will start making assumptions that it later agrees are incorrect assumptions.  When I give it prompts like "do not make assumptions about any conclusion, first ask me if something is true or not," it doesn't ask and continues the same habits.

&nbsp;

When I ask why it's doing that it concedes it did promise it would stop doing that (making assumptions and then spitting out advice on those assumptions) but then said it didn't stop and admits it continues to do it and therefore I (user) cannot rely on what it says or promises to do or not do.

&nbsp;

I can't be the only one this happens to.

&nbsp;

Default model used:  GPT-6.1 Sol Medium
```



document fetched 2026-10-28 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xgzes6/chatgpt_cant_or_wont_learn_from_its_mistakes
