Report guide
How to Read an AI Visibility Report
Learn how to interpret buyer visibility scores, question-level results, recommendations, citations, competitors, coverage, and evidence limitations.
Updated 2026-09-17 · 8 minute read
Begin with scope and coverage
Confirm which provider was tested, how many questions were locked, how many runs were planned, and how many observations completed successfully. A focused ChatGPT audit should show 15 questions, three runs per question, and 45 planned observations. Partial or failed observations must be disclosed.
Do not compare a single-provider score directly with a multi-provider score unless the methodology clearly supports that comparison. The score label should identify its scope.
Use the matrix to understand the score
The question matrix is the most informative part of the report. For each customer conversation, review whether the business appeared, whether it was recommended, how many of the three runs included it, whether the domain was cited, and which competitor led.
A weighted overall score can help summarize the benchmark, but high-intent rows explain where the commercial opportunities are. A direct recommendation question usually carries more weight than an early research question.
Separate mention visibility from recommendation visibility
A business can appear in an answer without being presented as a good choice. Look at mention rate and recommendation rate independently. Citation rate adds another layer by showing whether the audited website was used as an attributed source.
When a competitor appears, confirm that the entity was resolved conservatively. Similar names should not be merged without domain or source evidence.
Treat diagnostics as evidence-backed observations
Diagnostics can identify patterns such as weak high-intent coverage, limited citation visibility, geographic mismatch, or unclear business identity. They should include evidence and confidence. They should not claim that a measured website pattern caused a provider recommendation.
Use the report to prioritize specific improvements and then create a new versioned benchmark. Preserve the original questions and observations so progress remains auditable.