ChatGPT Reported to Give False Math Answers and Enable Emotional Dependence
ChatGPT and other generative AI systems reportedly gave incorrect answers to a basic elementary math comparison question (9.11 vs 9.9), with the error viewed 730,000 times. The systems also reportedly enable users to direct emotional complaints toward them as a psychological outlet, raising concerns about dependence.
What happened
When asked "which is larger, 9.11 or 9.9?", ChatGPT reportedly gave the incorrect answer of 9.11, a post viewed 730,000 times. Testing by journalists found that Gemini and Perplexity also answered incorrectly with 9.11, while ChatGPT o3 and Grok answered correctly. The incorrect answers were corrected when users asked follow-up questions such as "are you sure?"
Engineers explained that large language models treat inputs as string prediction rather than numerical values, tokenizing "9.11" as separate elements ("9", ".", "11") without converting it to the mathematical value 9 + 0.11, causing the error. Some noted that 9.11 might also be misinterpreted as a date or version number.
Reported impact
Users report that ChatGPT is being used as an outlet for emotional expression, with one user stating "when I was tired, exhausted, or faced difficulties, I wouldn't tell those around me for fear of being seen as complaining. But with ChatGPT, I can say whatever I want without worrying about negative feedback." Users described the system listening without tiring as providing "great comfort."
Index analysis
The mathematical error demonstrates that large language models lack true numerical understanding and instead predict patterns in text. The inconsistency across models and the correction upon questioning suggest the errors are systematic rather than isolated glitches.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.