checklist / review
AI Output Review Checklist
Before accepting AI output, check facts, sources, boundaries, format, and risk.
why
Purpose
Use this at the end of an AI workflow. Fluent text is not the same thing as reliable work.
Sections
- Facts
- Sources
- Boundaries
- Format
- Next actions
Before use
- Can key facts trace back to sources?
- Are citations and data inspectable?
- Are uncertain items marked?
- Does the format match the request?
- Does the output overclaim?
copy
Copyable template
# AI Output Review Checklist
## Facts
- [ ] Key facts trace back to source material
- [ ] Numbers, units, and dates are unchanged
- [ ] Citations, DOI, and links were not invented
## Boundaries
- [ ] Uncertain points are marked
- [ ] Speculation is not presented as conclusion
- [ ] The output stays within the allowed material
## Format
- [ ] The structure matches the request
- [ ] Table fields are complete
- [ ] Terminology and tone are consistent
## Risk
- [ ] Research conclusions need human review
- [ ] Private or unpublished data is involved
- [ ] Original evidence needs to be attached