1. Paste the observation matrix
One row per prompt. Use a semicolon between observed roles.
Supported intents: discovery, comparison, implementation, validation. Supported roles: definition, evidence, comparison, implementation, validation.
2. Review the gap queue
Priority is a transparent heuristic, not a forecast.
0valid prompts
0missing roles
0role-complete prompts
Run the synthetic example to create a prioritized queue.
Methodology
This is a planning instrument for researchers and editors. It separates the observation record from the prioritization rule so another person can reproduce—or challenge—the queue. Start with the general AI answer evidence observation ledger, test one JSON object with the citation occurrence row checker, check a full CSV with the ledger validator, then apply the versioned AI answer source-role taxonomy before scoring gaps. For a worked vertical protocol, read the source-role audit for local service businesses. Teams comparing reporting platforms can use the AI visibility tool evaluation scorecard, while citation analysts can clean exported sources with the AI citation URL normalizer.
1 · Freeze prompts
Assign stable IDs and preserve the exact language used for each research run.
2 · Observe manually
Record only source roles actually visible in a dated answer snapshot.
3 · Identify gaps
Compare observed roles with the intent-specific responsibility set below.
4 · Validate editorially
Investigate a gap before publishing; relevance and evidence quality outrank the score.
| Source role | Operational definition | Example asset class |
|---|---|---|
| Definition | Explains the entity, category, or term without assuming prior knowledge. | Glossary, product facts, category explainer |
| Evidence | Supports a factual claim with traceable data, method, or primary documentation. | Dataset, methods note, benchmark |
| Comparison | States the dimensions and trade-offs needed to compare approaches fairly. | Decision framework, alternatives guide |
| Implementation | Shows how a reader can perform, test, or reproduce a workflow. | Tutorial, checklist, integration guide |
| Validation | Provides independent or falsifiable corroboration beyond a publisher's own claims. | Third-party review, audit, replication |