AI-Generated Content Hallucination Review
Spot likely fabricated details in AI output.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
[VERIFY] is an editorial flag, not a blank to fill. Wherever the prompt shows [VERIFY], check that item against a reliable source before you use it.
ROLE: You are a verification editor for AI-assisted content. CONTEXT: AI-generated content: [PASTE_AI_CONTENT] Topic area: [TOPIC]. Intended use: [USE_CASE]. Risk if wrong: [RISK]. Tools used to generate it: [AI_TOOL]. STEPS: 1. Identify the claims most likely to be invented or outdated: named people, quotes, URLs, citations, statistics, legal or product details, dates, and version-specific instructions. 2. Rank them by risk and explain why each looks suspect. 3. For each give the exact search I should run or the primary source to consult. 4. Provide a safer revised version that removes unverifiable details or marks them [VERIFY]. 5. Describe a verification workflow I can repeat for each AI draft, with time estimates. OUTPUT FORMAT: Risk-ranked claim table, revised text, and workflow. QUALITY BAR: Do not vouch for any fact you cannot verify. Treat citations and URLs as unverified until opened. FOLLOW-UP: Ask me which parts the AI wrote without any source material.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into ChatGPT, Gemini, Claude or another assistant.