#Makes up information to tell you you're wrong

1 messages · Page 1 of 1 (latest)

hybrid treeBOT
#

Reported by @earnest parrot

Bug Report: Makes up information to tell you you're wrong
`Steps to Reproduce`

Make a correct factual claim. Ask it to very.

`Expected Result`

It will confirm the claim and possibly add some additional relevant detail.

`Actual Result`

It will respond with a hedged confirmation such as "largely correct, with caveats" or similar. Then it will refute some fact or detail that you did not say and that is the basis of it's "with caveats." It does this VERY CONSISTENTLY.

`Environment`

Chrome Browser, Windows 11, Plus account, Chat GPT 5.2 Instant

#
Additional Information

Please provide relevant details to help resolve the issue, such as:

  • ChatGPT Shared Link (if applicable).
  • Screenshots or videos demonstrating the problem.

-# ➜ Need to contact support? Visit the OpenAI Help Center.

earnest parrot
#

Here it literally admits to doing what the bug report is about, after I had to explain to it 1) what it did, and 2) that it's wrong. It's exhausting every time I try and use this thing to have to constantly correct it about not just what it's saying, but even it's reasoning, or approach.

#

And this is why I get so frustrated. Every substantive interaction with this tool is 80% me correcting it, or sorting out what it's conjured up in order to refute from the actual discussion.