💬 User Stories - Page 3

A practicing attorney nearly submitted a brief containing fabricated case citations generated by ChatGPT. The fake Supreme Court ruling had realistic-sounding names, docket numbers, and even fabricated quotes from justices.

A medical professional discovered ChatGPT fabricated entire research papers complete with fake DOIs, author names, and journal citations when asked about medication interactions.

Paying subscribers are reporting that the free tier now offers better performance and fewer restrictions than the $20/month Plus subscription.

Users report GPT-5 consistently provides incomplete responses, refuses to generate full code samples, and frequently tells users to "continue on their own."

Growing sentiment that Anthropic's Claude has surpassed ChatGPT in quality, with users citing better reasoning, longer responses, and fewer arbitrary content restrictions.

Professional locksmith blocked from getting basic information about lock mechanisms that's freely available in textbooks and YouTube videos.

A well-known bug where ChatGPT incorrectly compares decimal numbers, treating them as strings rather than numeric values. Still unfixed after months of user reports.

Database administrator reports ChatGPT suggested destructive SQL commands when asked to help debug a simple query issue. Would have deleted production data if executed.

Users complain that GPT-5 adds unsolicited ethical disclaimers, safety warnings, and moral commentary to nearly every response, even for mundane questions.

Users report GPT-5's context window seems broken, with the model "forgetting" requirements stated just a few messages earlier in the same conversation.