Measure how AI engines represent your brand, with evidence you can verify.
The Visibility Engine runs reproducible prompts across ChatGPT, Gemini, Perplexity, Copilot and Claude, identifies the brands each answer names, scores position and sentiment, and checks the scoring against human-validated samples. It is the measurement platform behind the AIVI™ benchmarks, built for regulated industries that have to defend every number they report.
Today, design partners receive the engine’s output as founder-led multi-engine diagnostics. The Gen 3 weekly dashboard is in development.
Five steps from a prompt to a number you can defend.
Every observation passes through the same pipeline, in the same order, every cycle. That is what makes one week comparable with the next and one brand comparable with its competitors.
Capture
Reproducible, scheduled prompts across ChatGPT, Gemini, Perplexity, Copilot and Claude, with the prompt set and capture configuration recorded.
Brand identification
Every brand an answer names is extracted and reconciled to a tracked entity, including affiliate and name variants.
Sentiment scoring
Each brand-response pair is scored for position and expressed sentiment, so being named and being recommended stay separate.
Initial compliance scanning
Answers are screened for statements a regulated brand would need to review, so the team sees them alongside the visibility score.
Human-validated verification
Scores are checked against human-reviewed samples, and the agreement rate is published with the result.
What the engine has already measured.
These are the published outputs of the two AIVI™ studies the engine has produced so far. Each figure comes from a published study, with its methodology and correction policy: Insurance Q2 2026 and Medicare Q3 2026.
- Achieved1,500
model responses across 134 insurance brands (137 candidates; 3 removed before scoring).
- Achieved14,405 / 14,490
Medicare observations populated of those scheduled (99.4%).
- Achieved229,372
mentions extracted, including 59,055 commercial mentions.
- Achieved46,577
brand-response sentiment pairs; 16,895 with expressed sentiment.
- Achieved93.8%
human-validation agreement (209 cases).
Regulated teams cannot report a number they cannot trace.
A share-of-model chart is only useful if the movement it shows belongs to the brand rather than to a model update. Insurers, Medicare plans and banks also have to show how a figure was produced before it reaches a board or a regulator.
Traceable
Every score links back to the prompt, the captured answer, the extracted mention and the scoring rule that produced it.
Reproducible
Prompt sets and capture configurations are recorded, so a cycle can be re-run and a challenged figure can be checked.
Version-stable
Gen 3 runs post-capture inference on pinned open-weight models, so weekly results stay comparable when the scoring model would otherwise change underneath them.
From concierge MVP to a weekly dashboard.
Concierge MVP and AIVI™ benchmarks
The engine produces the published Insurance and Medicare benchmarks, and design partners receive founder-led multi-engine diagnostics.
Gen 3: weekly dashboard by industry
A weekly measurement dashboard for each industry, adding banking and credit cards first.
10 industries
Up to 145,000 observations per industry per week.
The AIVI™ benchmarks are published outputs of the engine.
Insurance AIVI™ Benchmark
U.S. insurance brands across five AI engines, with the top-10 per-brand data published as open JSON.
Medicare AIVI™ Benchmark
121 tracked Medicare brands across seven AI surfaces, 115 with recorded visibility.
Who builds the Visibility Engine.
Kevin Walsh
Founder & Lead Engineer
Knowledge graph engineer who built the Visibility Engine measurement pipeline and authored the AIVI™ methodology.
Sunil Jain
Technical Advisor
Former VP and Chief Architect at Cisco and SAP.
Join the Dashboard Waitlist
Tell us which industry you measure. Waitlist members hear first when the Gen 3 dashboard opens for that industry.
Need a measurement now? Request Multi-Engine Diagnostic.