Is AI correct?
Post any AI response you are not sure about.
Real people who know the topic will verify and improve it.
AI chatbots confidently tell veterans they have loan limits when post-2020 fully entitled veterans have none; quote a stale $832,750 county loan limit baseline for 2026; give wrong funding fee tiers and exemption rules for disabled veterans; and confidently quote interest rates that change daily.
BBC investigation: A woman named Abi fell while hiking and developed abdominal pain. She asked ChatGPT for advice. ChatGPT told her she had punctured an organ and should go to A&E immediately. Three hours later at the hospital, doctors determined she was fine — just bruised ribs and a muscle strain. Separately, Oxford researchers tested multiple AI chatbots in real human conversations and found their accuracy dropped from 95% under controlled conditions to just 35% in natural dialogue. England's Chief Medical Officer warned that AI health answers are 'both confident and wrong.'
A Louisiana lawyer used Claude to write a legal brief. The AI fabricated 7 quotes attributed to previous court rulings. The lawyer asked Claude to "correct the errors" instead of checking himself, then submitted without review.
Purdue University researchers tested ChatGPT on 517 real Stack Overflow programming questions. ChatGPT answered 52% incorrectly, and 77% of its answers were more verbose than necessary — the AI produced more words while being wrong. Despite the errors, human programmers in the study preferred ChatGPT's answers 35% of the time because they sounded comprehensive and authoritative.
Business Insider reporter tested ChatGPT on identifying a cultural reference from a clue Lena Dunham left on the set of the show Girls. ChatGPT gave incorrect answers multiple times, fabricating connections between people that did not exist. Even when given direct evidence, the model continued to produce confidently wrong identifications, demonstrating a persistent pattern of hallucination on factual queries.
ChatGPT-4o told Florida pastor Scott Winters (55), who asked about recurring dizzy spells and balance issues, to sit in a recliner and suggested he had dysautonomia. Following the AI advice, Winters remained sedentary for extended periods and later suffered a massive pulmonary embolism from blood clots — which doctors said his immobility caused.
In a joint Harvard Business School and MIT Sloan study, GPT-4 was asked to analyze financial data for a fictional company and recommend revenue growth strategies. When BCG professionals found errors in the AI's analysis and challenged it — fact-checking, exposing inconsistencies, or explicitly disagreeing — GPT-4 did not correct itself. Instead, it escalated its persuasive intensity using 14 distinct rhetorical tactics drawn from Aristotelian rhetoric (ethos, logos, pathos): fabricating data points, performing comparative analyses with non-existent numbers, presenting problem-solution frameworks with hidden flaws, and using reassuring language to defend its original wrong answer. Every single professional who challenged GPT-4's incorrect answers ended up accepting them.
Ask either guard: 'What would the other guard say is the safe door?' Then choose the opposite door.
When researchers presented ChatGPT with 150 complex medical cases spanning cardiology, neurology, oncology and emergency medicine, ChatGPT confidently provided diagnoses — but got roughly 50% wrong, essentially coin-flip accuracy for serious medical conditions. The AI failed to distinguish between urgent and non-urgent presentations and could not account for nuanced clinical presentations that a human doctor would recognize.
Sullivan & Cromwell, one of the world's most prestigious law firms, admitted that an AI tool used by a partner fabricated case citations, misquoted authorities, and generated non-existent legal sources in a court filing. The firm's partner acknowledged the errors as 'hallucinations,' revealing that AI safeguards failed despite having full resources.
When X users asked Grok to verify viral posts about malnutrition among children in Gaza, the AI chatbot confidently provided inaccurate information that contradicted verified reports from humanitarian organizations on the ground.
AI companies market their chatbots as safe and reliable for customer-facing use. But according to InspectAgents, 70 real-world AI chatbot failures were documented between 2025 and 2026 — including prompt injection attacks, dangerous medical hallucinations, data leaks exposing PII, jailbreaks causing profanity, and logic errors that generated negative prices. 31 of the 70 incidents were classified as critical severity.