Checklist / Research workflows / Intermediate
AI Research Quality Checklist
Run this before an automated research output is shown to anyone who will act on it. It is a review aid, not a certification.
For AI engineers, research teams
01Plan
- The research question is written down and narrow enough to answer
- Sub-questions are listed separately rather than bundled
- The intended source types are named before searching begins
- An acceptable maximum source age is defined per claim type
02Sources
- Every source URL is recorded with its retrieval timestamp
- At least one independent source supports each material claim
- Vendor claims are labelled as vendor claims, not as findings
- Sources that were planned but unreachable are listed, not dropped
03Claims
- Each sentence that states a fact carries a citation
- Each citation carries a supporting excerpt, not only a URL
- Conflicting sources are presented as conflict rather than resolved silently
- Inferences are visibly separated from sourced statements
04Handover
- The output states what was searched and what was not found
- Coverage is reported as sources reached versus sources intended
- A reviewer can verify any single claim in under thirty seconds
researchcitationsquality