How we measure
How we measure
AI visibility
Every score, grade, and dataset on this site comes from one method: asking real buyer-intent questions to live AI engines and reading the verbatim answers. Here is exactly how it works, what the numbers mean, and where the method has honest limits.
On this page
In one line
AI visibility score = % of tested prompt×engine runs where the brand was mentioned, against live engines, averaged over multiple runs. Grades: invisible, barely visible, partially visible, highly visible.
What we measure
AI visibility is a measure of what generative engines actually do with your brand, not what they might do. Every scan records five things:
- Mention rate. The share of tested prompt×engine runs where the business is named, by name or domain.
- Position. How high the first mention appears in the answer — first is a recommendation, deep is a footnote.
- Sentiment. The tone of the context around the mention, classified from the surrounding text.
- Citations. The domains the engine actually cited, deduplicated and counted — this shows the sources AI trusts instead of you.
- Coverage. Google AI Overview presence and GA4 referral sessions, tracked separately on higher plans.
The mention rate is the primary score; everything else explains why it moved.
How a scan runs
Each scan is a matrix of prompts × engines. In this order:
- Prompts. We generate buyer-intent questions relevant to the business’s category — up to 15 per location, tacked to real discovery language (“best X near me”, “what’s the best-rated X in …”), never containing the brand name so we measure organic recommendation, not self-quoting.
- Engines. Every prompt runs against the configured engines with live web access on the same day, so answers reflect current sources.
- Mention detection. Each answer is matched against the business’s name and domain to decide whether it was mentioned, and where.
- Enrichment. Sentiment is classified on mention-bearing answers; cited domains are extracted and deduplicated.
The free public check is a compressed version: three prompts across three zero-cost engines, returning score, grade, mention status, and per-engine positions.
Scoring & grades
The AI visibility score is the percentage of successful prompt×engine tests that mentioned the business:
- 100/100 — named in every tested answer.
- 0/100 — named in none. Invisible to AI.
Errored or refused engine calls are excluded from the denominator so a cold model never drags the score. Grades map to bands so one-off LLM wobble never flips a grade:
- 0 → Invisible · never named
- 1–39 → Barely visible · named only occasionally
- 40–74 → Partially visible · named with middling consistency
- 75–100 → Highly visible · reliably named and usually recommended
Data quality & limits
We are honest about what AI measurement can and cannot say:
- Non-determinism is real. Generative engines produce different answers run to run. Multi-run and multi-prompt averaging dampens this; grade bands absorb the residue.
- Sampling, not census. A scan tests a sample of the infinite prompt-space. It is directionally exact and repeatable, not an exhaustive rank.
- Refusals are excluded. When an engine declines to answer, the test isn’t counted as a miss.
- Scores are comparative over time. Read them as a slope: your score versus your own history, and versus competitors scanned on the same day with the same prompts.
Run the same method yourself
The measurement is not proprietary guesswork — it is ask-the-engine audit. Run it on any brand in a minute with the free AI visibility checker, browse category benchmarks, or read the public leaderboards for opted-in brands. For the playbook behind the numbers, start with the AI visibility guide.
How is the AI visibility score calculated?
We run real buyer-intent prompts against live AI engines, analyze each verbatim answer for a mention of the business, and compute the score as the percentage of successful prompt×engine tests in which the brand was mentioned. A score of 100 means every tested answer named the business; 0 means none did.
Which AI engines do you scan?
The free check runs three live models at zero cost through OpenRouter, Groq, and Mistral. Paid plans add ChatGPT, Gemini, Claude, and Perplexity, each with real web access, plus separate tracking for Google AI Overview coverage and GA4 referral sessions.
What does a 'mention' mean exactly?
A mention is a detected reference to the business by its name or domain in an engine's answer. We also record the position of the first mention (how high in the answer the brand appears), classify the sentiment of the surrounding text, and capture the domains the engine actually cited.
Can AI scores fluctuate between runs?
Yes — generative engines are non-deterministic, so a single answer can vary. Where variance matters we average results across multiple runs and multiple prompts so a one-off answer does not get treated as real movement. Grade bands (invisible, barely visible, partially visible, highly visible) are wide enough that small wobbles never flip a grade.
What do the grades mean?
A score of 0 is 'invisible' (never named), under 40 is 'barely visible', 40–74 is 'partially visible', and 75–100 is 'highly visible'. The transition out of 'invisible' is typically the fastest — it usually only requires entity clarity and content that directly answers the tested questions.
How often should I scan?
Weekly is the sweet spot for detecting real change without noise. Free plans scan weekly; Starter and above scan daily; Growth and Agency plans add real-time scanning with alerts the moment a competitor displaces you.
Is AI visibility the same as Google ranking?
No. Google ranks pages in search results; AI engines synthesize answers from entities, citations, and sources. You can rank #1 in Google and remain invisible to ChatGPT, and vice versa. We measure both as separate surfaces because they need separate fixes.
Where does the public dataset come from?
The 2026 AI visibility dataset aggregates mention rates and scores by vertical and region from every scan run on the platform — opted-in brands only, aggregated so no individual business is identified. It updates automatically as scans complete.
Related reading
See your brands score today
The free check runs the same prompt×engine matrix described above — score, grade, and the competitors taking your slot — in under a minute. No account needed.