Define the measurement decision
Use within-panel comparisons first. Cross-company or industry claims need consistent collection and disclosed exclusions.
Measurement method
| Step | Action |
|---|---|
| 1 | Freeze the benchmark prompt set and publish its scope |
| 2 | Count scheduled, successful, failed, and excluded observations |
| 3 | Calculate brand coverage within successful responses using a stated denominator |
| 4 | Compare position, Sentiment, and citations as separate dimensions |
| 5 | Show uncertainty and avoid ranking tiny differences as meaningful |
Worked calculation or observation
If Brand A appears in 12 of 20 successful purchase prompts and Brand B in 11, the difference may not justify a strategic claim. Reading the answers and sources may reveal a more actionable distinction.
Evidence to retain for brand-visibility benchmarking in AI answers
Keep these fields with the decision:
- metric name
- numerator
- denominator
- scope
- platform
- Prompt Group
- success state
- observation date
- formula version
Limits and UnderAI's role in brand-visibility benchmarking in AI answers
Benchmarks cannot be generalized beyond their prompts, dates, platforms, and markets. They also do not prove future performance.
UnderAI Platform Overview and snapshots provide scoped data; public benchmark pages should expose methodology and never combine unsupported dimensions into one score.