Plain AI in plain English
About

Trust & truth · the practice

Cross-checking AI

30-second gist~30s read

For anything that matters — money, medicine, law, dates, names, numbers — don't trust the AI's answer alone. Cross-check is the habit that catches almost everything that's wrong.

It's quick. Three steps. Worth doing.

If you want more

The 30-second cross-check3 steps
  • 1. Search the key claim. Take the most specific fact in the answer (a number, a date, a name) and Google it. If you find the same claim repeated by independent, named sources, it's probably real. If you only find AI-shaped websites repeating it, it's probably not.
  • 2. Click any citation. If the AI mentions a source, open it. If the link doesn't load, the source doesn't exist, or the page doesn't actually say what the AI claimed — that's your answer.
  • 3. Ask the AI: "what's the strongest reason this might be wrong?" A good answer breaks down. A bad answer holds up.
What never to skip the check on~30s
  • Medical doses, drug interactions, "is it safe to take X with Y?"
  • Legal advice, contracts, statutes.
  • Financial decisions, tax rules, bank-account moves.
  • Anything tied to a date, a price, or a name.
  • Anything you'd quote in writing, especially professionally.