Methodology & definitions
How we measure AI visibility
AI visibility should be measured like a survey, not a screenshot. Answerly samples each prompt repeatedly across ChatGPT, Gemini, Perplexity and Google AI Overviews (Google AI Mode is in progress). and reports every rate with its sample size (n) and a 95% margin of error. A change smaller than the error bar is shown as noise, not progress; each surface is measured separately; and the raw answer behind every result is stored so any number can be verified.
Repeated sampling, not one answer
AI answers are non-deterministic — the same prompt can name different brands on different runs. So a metric comes from many runs of a prompt, not a single reply. A screenshot of one good answer proves that the answer happened once, which is not the same as knowing how often it happens.
Sample size and margin of error
Every rate is shown with n and a 95% margin of error, and comparisons are weighted by sample size so a small spot check cannot overwrite a large baseline. A movement inside the error bar is labeled noise. We would rather report a boring true number than an exciting one we cannot defend.
Mention, citation, recommendation, share of voice
A mention is your brand named in the answer text. A citation is your site linked as a source. A recommendation is the assistant suggesting you. Share of voice is your mentions relative to competitors across the set. Four distinct outcomes, tracked as four metrics — not folded into one score.
Per-engine, never a blended rank
ChatGPT, Gemini, Perplexity and AI Overviews retrieve and cite differently, so results are reported per engine. There is no single cross-engine "AI rank", and inventing one would average away the very differences you need in order to act.
Unverifiable entities are flagged
Assistants sometimes return names with no real site behind them — surfaced through a maps link or a thin listing. These are flagged as unverifiable and kept out of confirmed-competitor metrics until reviewed, so share of voice is not inflated by entities nobody can check.
Evidence is stored, not summarized away
The raw answer behind every mention and citation is kept with its engine and date, and can be opened as a prooflink. That is what makes the metrics auditable: any number in a report can be traced back to the dialogue that produced it.
What we do not claim
We do not promise placement in AI answers, we do not report engines we cannot actually measure, and we do not present a single run as a trend. Where coverage is partial — such as source lists that some engines do not expose — the gap is reported as a gap.
Related
FAQ
How should AI visibility be measured reliably?
By sampling each prompt repeatedly across engines and reporting mention and citation rates with sample size and a margin of error — not by screenshotting one answer.
What sample size do I need?
Enough runs that the margin of error is small enough to decide on. Answerly always shows n, so a spot check is never mistaken for a measurement.
What is the difference between mention rate, citation rate and share of voice?
Mention rate is how often your brand is named; citation rate is how often your site is linked as a source; share of voice is your mentions relative to competitors across the prompt set.
Why not report one combined AI visibility score?
Because engines behave differently, a blended score hides where you actually win or lose. Answerly reports per engine and keeps the underlying metrics separate.
See if AI recommends your brand
Start free — no card. Run your first AI visibility check across ChatGPT, Gemini and Perplexity.