Skip to content
Automation Squad
Hand to your teamWritten for Any two of ChatGPT, Claude and Gemini

Have one AI audit another one's answer before I trust it

Robert MacKelfresh

By Robert MacKelfresh

Founder, Automation Squad

Two models trained differently fail differently, so the second one catches what the first was confident and wrong about — which is exactly the failure a single model cannot self-report. Section 2 is the whole point: a fabricated specific sits inside otherwise-correct text at the same confidence level, and it is nearly invisible unless something is explicitly hunting for it.

The prompt

Below is an answer produced by a different AI to the question: [THE ORIGINAL QUESTION YOU ASKED]

You are the auditor. You have no loyalty to it and no obligation to be polite about it. Do not rewrite it — I'll decide that.

Go through it and mark up:

1. CLAIMS OF FACT — pull each one out as a separate line. For each, mark it: VERIFIED (you're confident and can say why), UNVERIFIABLE (might be true, nothing supports it here), or WRONG (say what's actually the case).

2. THE CONFIDENT GUESS. Somewhere in a long answer there's usually a specific detail — a number, a name, a date, a version — stated with the same confidence as the well-supported parts but which is actually a reconstruction. Find it. This is the single most useful thing you can do here.

3. WHAT IT LEFT OUT. Anything a knowledgeable person would consider a serious omission given the original question.

4. WHERE IT ANSWERED A DIFFERENT QUESTION. Models drift toward the question they can answer well. Point at any place that happened.

5. YOUR VERDICT, in one line: use as-is · use with the fixes above · start again.

Do not add new content of your own beyond what's needed to make a correction. If you think the answer is genuinely good, say so — but you have to have done sections 1 to 4 first.

---
THE ANSWER TO AUDIT:

[paste the other AI's full answer here]

One more thing: use anything you already know about me from our past chats and anything you're connected to. If you can find it yourself, don't make me look it up.

Copy it, paste it, run it. Nothing here is gated.

More for the same job

Written for Any two of ChatGPT, Claude and Gemini as of August 2026. Models change — if this stops working the way it should, that is worth telling us about.