AI visibility
AI visibility is how often a brand appears in the answers AI assistants give to relevant questions, measured as the share of sampled answers that mention it.
Last updated September 28, 2026
How it is measured
The method is to write a fixed set of benchmark questions, ask each AI provider those questions, and record whether the brand appears in each answer. Visibility is the number of answers in which the brand appears divided by the number of answers sampled. Questions should be unbranded, such as "which US airlines are the most reliable?" rather than "is Delta reliable?", because a branded question guarantees a mention.
Two further records make the number useful. Which domains each answer cited shows what the models relied on, and a breakdown by provider shows whether the brand appears in one assistant and not another. Visibility is a presence measure. It does not say how the brand was described, and describing tone requires separate analysis.
Variance is large. Live answers change between runs, and one brand moved from 40% to 30% to 70% across three runs of the same questions in PeakMetrics testing. Small samples make single numbers fragile, so AI visibility should be reported as a dated snapshot with its sample size, and repeated before it is treated as a trend.
What PeakMetrics data shows
PeakMetrics' AI Perceptions free check compares AI visibility with 30 days of press coverage from the PeakMetrics news index. Measured on 29 September 2026 with ChatGPT, Claude, Gemini, Perplexity and Grok (1,250 answers per sector; scope and caveats on the free check page), two patterns stand out.
First, press volume and AI visibility diverge. In the airlines sector, Southwest Airlines was named in 575 of 1250 answers (46%, range 43–49%) with 7,951 news mentions in 30 days. American Airlines was named in 466 of 1250 (37%, range 35–40%) with 32,472 news mentions, about four times as many. Delta Air Lines was named in 843 of 1250, with 14,417. The two ranges do not overlap, so the gap is not an artefact of sample size, but one run cannot say why the measures separate.
Second, some brands with heavy coverage did not appear at all. Of the 298 brands measured across 20 sectors, 3 appeared in none of the 1,250 answers collected for their sector. Among them, BlackRock had 64,484 news mentions, SiriusXM 8,269 and Flexport 191 in the 30 days to 29 September 2026. A zero here means the brand was not named in those 1,250 answers to those category questions. It does not mean the brand is invisible to AI generally, and it says nothing about how any of these companies is regarded.
Provider differences are also visible. Delta appeared in 180 of 250 ChatGPT answers but 134 of 250 Perplexity answers, which is why sampling one assistant can mislead.
Why it matters to communications and risk teams
AI visibility tells a team whether the brand is part of the answer when the public asks about its category. It complements media monitoring, which reports what was published, and is the outcome that generative engine optimization work targets. The cited domains show which publishers to prioritize, and include wire and newsroom content (see press release syndication).
Common misconceptions
AI visibility is not share of voice. Share of voice divides a fixed pool of mentions among competitors, whereas visibility counts appearances in sampled answers, and several brands can each be visible in the same answer. It is also not a favorability score, and it is not stable: treat it as a dated measurement. And the PeakMetrics AI Perception Score is a separate, defined metric in the paid product; the free check reports visibility only.
See how these signals show up in your own coverage on the PeakMetrics platform or run the free AI Perceptions check. Back to the glossary.