Thinking
critical-thinkingverification
Updated June 2026
Scrutinize a claim or argument, separating what's verifiable from what's a confident guess.
Why it works: Forcing a split between verified, inferred, and unknown stops the model from laundering a recollection into a stated fact.
This is a message template. Fill in the blanks below, then paste it into the chat.
Scrutinize the following before I rely on it. Be honest about the difference between what you can verify and what you’re pattern-matching.
The claim: [paste the claim, argument, or passage]
Where it came from: [source, if you have one, or “unknown”]
How I plan to use it: [context, so you know the cost of being wrong]
Please:
- Break it into the individual factual claims it depends on
- Label each as verifiable, plausible-but-unverified, or likely-wrong, with your confidence and what it rests on
- Flag the parts most prone to confabulation: specific numbers, dates, quotes, version strings
- Tell me what evidence would settle each one, and where I’d look
Don’t manufacture a citation. If something is checkable and you can’t check it, say so plainly and tell me how I can.