# Help me understand ChatGPT's limits. — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/c3181ac15eb9fc0e

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-09-24 17:14 ingested 2026-09-25 02:37

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1wp7k10/help_me_understand_chatgpts_limits/)

## retained text



```
I understand what an LLM is and why this isn't some sci-fi AI system. However, there are basic tasks I would suspect should be in scope: asking it to essentially google search some info for me and take the answer from the most authoritative source, usually this is a very controlled primary source on a government regulation website. 

I could do it myself, obviously, but I would like to use ChatGPT to save some time. Luckily, I know what I'm doing and can usually tell instantly when it gives me a wrong or outdated answer.

ChatGPT's response to why it gets things wrong usually comes down to: Sorry, I didn't search rigously enough or from the correct primary source and instead went off of old data and didn't do any verification checks.

Even if I ask it to always triple check and cross reference and present conflicting information and defer to me instead of just deciding for itself...it still constantly makes these mistakes. 

This is a business account using the latest model...


I would probably trust a 10 year old to make these google searches sooner than trust ChatGPT.
```



document fetched 2026-09-25 00:21 · canonical https://reddit.com/r/ChatGPT/comments/1wp7k10/help_me_understand_chatgpts_limits
