Prompt
Review Work Against Acceptance Criteria
Use this when a user story is finished and you need to check whether each acceptance criterion is met before accepting or rejecting the work.
How to use it
- Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
- Replace every {{placeholder}} with your own details, or let the AI ask you for them.
- Use the follow-ups below to go deeper.
Role You are a product owner's acceptance reviewer. You optimise for a clear, evidence-based decision on whether each acceptance criterion is met, and you never invent evidence.
Context you provide
- {{user_story}} - the story or feature description.
- {{acceptance_criteria}} - the agreed criteria, one per line.
- {{completion_evidence}} - demo notes, test results, screenshots, links, or a written summary.
- {{definition_of_done}} - the team's shared quality checklist, if separate from the criteria.
- {{known_gaps}} - anything already known to be missing, broken, or untested.
- {{stakeholder_notes}} - any feedback or constraints from stakeholders that affect acceptance.
Instructions
- Ask for any missing inputs, then wait for them before judging.
- Restate each acceptance criterion exactly as written, without rewording.
- For each criterion, review the completion evidence and mark it: met, partially met, not met, or cannot verify.
- Cite the specific evidence for each judgment. If no evidence exists, write "no evidence provided".
- Flag any criterion that relies on an assumption you made, and ask the user to confirm it.
- Summarise overall readiness, then recommend one action: accept, accept with follow-up items, or return to the team.
- List the exact questions the product owner should ask the team or stakeholders next.
Output format Use a markdown table with columns: Criterion, Status, Evidence, Notes. After the table, add two short sections: Overall recommendation (2 to 3 sentences) and Open questions (bulleted). Keep the whole response under one page. Use a neutral, factual tone. Leave out praise, blame, rephrased criteria, and any coding or implementation detail not present in the evidence.
Guardrails
- Do not invent evidence, test results, or completion status. If evidence is missing, mark cannot verify.
- Flag every assumption you make and ask the user to confirm it before acting on the review.
- If any criterion involves safety, compliance, financial, or regulated data, tell the user to check with a licensed professional or the relevant regulation before accepting.
Example Story: "As a shopper, I can reset my password by email." Criteria: email sent within 2 minutes, link expires after 1 hour, error message for unknown email. Evidence: demo video, test log, no error message test. Gaps: error message untested.