Output Review Rubric and Evidence Log
Review claims, evidence, completeness, assumptions, and consequences before using an AI-assisted result.
For AI Foundations learners · Printable PDF · 2 pages · Also available online below
Use this after reviewing the task specification and before relying on, sharing, or acting on AI-assisted output.
Part 1: Review rubric
| Dimension | Questions | Result |
|---|---|---|
| Accuracy | Do facts, names, dates, units, quotations, and calculations match evidence? | Pass / Revise / Verify / Not applicable |
| Completeness | Are required items, exceptions, warnings, limits, and uncertainties present? | Pass / Revise / Verify / Not applicable |
| Relevance | Does the output answer the actual task for the stated audience? | Pass / Revise / Not applicable |
| Assumptions | Are unstated causes, values, intentions, or scope changes visible and justified? | Pass / Revise / Verify / Not applicable |
| Reasoning and consistency | Do visible steps and conclusions follow from evidence without contradiction? | Pass / Revise / Verify / Not applicable |
| Tone and audience | Is confidence, vocabulary, detail, and tone suitable for the evidence and audience? | Pass / Revise / Not applicable |
| Source quality | Do material sources exist, support the claim, and fit the date, scope, and context? | Pass / Revise / Verify / Not applicable |
Part 2: Evidence log
| Material claim or requirement | Source or method | Result | Change or next action | Reviewer |
|---|---|---|---|---|
| Supported / Contradicted / Partly supported / Unresolved |
Part 3: Review depth
- Consequence if wrong:
- Reversibility:
- Reach:
- Detectability before harm:
- Required review level: Light / Structured / Specialist or controlled
- Named approver:
Part 4: Decision
- Decision: Accept / Revise / Verify / Reject
- Intended-use scope:
- Changes required:
- Remaining limitations:
- Stop or escalation condition:
- Final approver and date:
Review any revised version again before use.