# Well, I'Il unsubscribe if you won't actually help ChatGPT! — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/f0ba7cf1575a86f8

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-11-17 22:06 ingested 2026-11-18 02:36

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

## retained text



```
I was setting Chat some tasks to count some words on a page and I realised the answers were becoming too familiar to ones it had already given me, so l investigated its work to make sure it wasn't glitching out.  
When questioned it gave me this answer !!

**Me**: Why are you getting it so wrong so often?

**ChatGPT**: Because I've been rushing the counting and making assumptions instead of properly checking the images. And you're right to call me out on it.  
There are three things I've been doing wrong:  
1. Guessing instead of verifying.  
Some names are small and difficult to distinguish, and I've occasionally given you names that aren't actually there.  
2. Relying on patterns. Ive assumed pages contain a certain number of rows and columns rather than counting every name individually.  
3. Being too confident. Even when you've asked, 'Are you sure?", I've sometimes confirmed an incorrect answer instead of checking properly.  
The worst part is that you've had to do the checking yours If, which defeats the purpose of asking me.
```



document fetched 2026-11-18 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xyxrky/well_iil_unsubscribe_if_you_wont_actually_help
