Methodology
How MentionAudit measures AI-search visibility.
We run the same buyer-intent prompts across the major AI-search engines, keep the raw evidence, and score whether your brand was cited, merely mentioned, or skipped. That is the whole system.
v1.0·Last updated→What we test
MentionAudit runs a fixed set of buyer-intent prompts across ChatGPT, Claude, Gemini, Perplexity, and Grok. Every engine gets the same question set, region, and prompt version so the results are comparable.
- ChatGPT OpenAI
- Claude Anthropic
- Gemini Google
- Perplexity Perplexity AI
- Grok xAI
What we store
We store the raw answer, exposed source URLs, model identifier, timestamp, region, and prompt version. We do not retry until an answer looks flattering; the first usable response is the response we score.
How we score
The score answers one buyer question: did the engine recommend your brand, cite it as a source, only mention it, or skip it?
- Citations count more than mentions because attribution is the durable visibility signal.
- Engines that skip the brand count as misses instead of disappearing from the average.
- Recommendations must trace back to captured answer text or source evidence.
Methodology version history
- v1.0 · 2026-05-08Beta launch — published engine coverage, citation detection, crawl × cite ratio, boundaries, and the audit-detail version banner.
- v0.9 · 2026-04-22Added closed-loop recommendation verification (Pro Brand and Agency) — re-audits the queries a recommendation references and reports before/after deltas.
- v0.8 · 2026-04-11Added Grok coverage and the Cross-Engine Patterns module that surfaces shared citation domains across covered engines.
- v0.7 · 2026-03-30Added Pass 1.75 Engine Diagnosis — every audit run records why a given engine cited or skipped the brand instead of inferring from totals.
- v0.6 · 2026-02-15Pro Brand correlation panel — joined LLM citations to AI-crawler hits per engine and per day to surface the citation gap.
What we do not cover
These are scope boundaries, not hidden fine print. We publish them so the score is clear about what it measures.
- Personalized or logged-in answer paths, including custom memories and private GPTs.
- Competitor-side crawler traffic unless the customer owns the measured property.
- Engines outside the current covered set, such as You.com and Brave Search.
- Minute-by-minute answer drift after the audit snapshot has been captured.
Request removal from public proof material
If your brand appears in public MentionAudit proof material and you want it removed, submit the form below. We respond within 5 business days and remove the brand from internal observational benchmarks within 30 calendar days.
Run the same capture
See where your brand shows up.
Submit a domain to see the same engine set, citation detection, and methodology version on your own audit.