HUMAN REVIEW SELF-CHECK — RUN THIS PROMPT ON AI OUTPUT BEFORE YOU USE IT Zero → Frontier — free, ungated, no attribution required HOW TO USE THIS: After an AI tool drafts something for you, paste that draft BACK into the same or a different AI tool with this prompt. You're using the AI to help audit its own (or another model's) output before you trust it — this does not replace your own judgment, it structures it. --- I want you to review this AI-generated output critically, as if you were checking someone else's work, not defending it: OUTPUT TO REVIEW: [paste the AI-generated draft/output] WHAT IT WAS SUPPOSED TO ACCOMPLISH: [state the original goal] WHERE IT WILL BE USED: [e.g. "sent to a client," "published publicly," "internal only"] Check it against these seven points and flag anything that fails, rather than assuming it passes: 1. FACTUAL CLAIMS — does it contain any number, date, name, or statistic I should independently verify rather than assume is correct? 2. ATTRIBUTION — does it attribute anything to a specific person, company, or study? If so, flag it as needing verification. 3. TONE AND VOICE — does this sound generic/AI-written rather than like an actual person wrote it? Point out specific phrases that feel like AI filler or hedging. 4. IRREVERSIBLE ACTIONS — does using this output as-is trigger anything that can't be undone (a send, a payment, a public post, a legal or medical claim)? If so, say explicitly that a human must confirm before that action happens. 5. PRIVACY — does this reveal or reference anything that looks like confidential, proprietary, or personal information that shouldn't be here? 6. BIAS OR FAIRNESS — is there any stereotype, unfounded assumption, or exclusionary language, even subtle? 7. FINAL VERDICT — would you be comfortable if a human put their name on this exactly as written? If not, say specifically what needs to change before it should go out. Be direct about problems — don't soften a real issue into a vague suggestion. --- Free to copy, adapt, and reuse. Part of the Zero → Frontier project.