Check an AI answer

AI assistants can sound sure and still be wrong. Run through these seven checks before you use an answer for anything that matters. Tick them as you go; nothing is saved.

Version 1.1 · October 26, 2026 · What’s new

A prompt that asks the assistant to check its own answer, a reset button, and a progress count.

Is the main claim true? Search for it in a source you trust. If you can't find it anywhere, treat it as unconfirmed.
Are the sources real? Open every link and reference. Assistants sometimes invent titles, authors and URLs.
Do the numbers add up? Recalculate totals, percentages and conversions yourself.
Are dates and names right? Check people, places, product names and dates against an official page.
Is it current? Ask yourself when the model's knowledge ends, and whether anything changed since.
Is anything missing or one-sided? Ask the assistant for the strongest counter-argument, then judge.
Would you sign it? Read it aloud. If it doesn't sound like you or makes promises you can't keep, rewrite it.

Ask the assistant to check itself

Paste this right after an answer you want to use. It doesn’t replace your own checks, but it surfaces weak spots fast.

Check your last answer before I use it. For each factual claim, number or date: say whether you are sure, and how I can verify it (a source I can open). Flag anything you may have guessed or invented. Then give me the strongest reason your answer could be wrong.

Learn more: Why can't you trust everything ChatGPT says?

Check an AI answer

Ask the assistant to check itself

Get the free Guide to Artificial Intelligence