Spot wrong answers

VerifyDrill trained your detective eye: look for checkable claims, fuzzy sources, math that doesn’t add, and advice that ignores your constraints. You practiced not trusting tone. Fluency is a costume; verification is the flashlight.

A fast verify loop:

1. Underline claims you could prove or disprove (numbers, names, dates, “everyone agrees”).
2. Ask for sources or steps — then check them outside the chat.
3. Re-ask with a trap: “What might be wrong in your answer?” or “List assumptions you made.”
4. Cross-check one important fact with a trusted site, textbook, or person.

Verification recipes:

Review your previous answer.
1) List every specific fact (numbers, names, dates, product claims).
2) Mark each: Confident / Unsure / Likely guessed.
3) For anything Unsure or Likely guessed, tell me how I should verify it in under 2 minutes.

I cooked with your recipe and the chicken was underdone at 18 minutes.
Given: bone-in thighs, skillet, medium heat.
Revise times/temps carefully. Say what you might have gotten wrong.
Do not invent food-safety advice beyond common home-kitchen practice — remind me to check with a reliable source if unsure.

Study version: “Quiz me, then show which of your earlier explanations conflict with my textbook page [X].” Work version: “Which deadlines did you invent vs quote from my notes?”

• Health-ish or money-adjacent tips (then leave chat and verify properly)
• Citations, statistics, and “recent news” claims
• Recipes, travel times, and anything with a number

Asking the same chat “are you sure?” often gets a polite double-down. Prefer an outside check. Never paste private medical records to “verify.” For harm-sensitive topics, human professionals beat chat polish.

Grab any recent AI answer. Underline three claims. Verify one online or against your notes. Note what failed.