Method
How to Measure AI Visibility: Sections Read and Cited
AI visibility is not one number. It comes from comparing which sections of a page Google indexes with the sections that ChatGPT, Perplexity, Gemini, Grok and Google AI actually cite.
If you want to know how AI visibility is measured, start by separating two readings that are often mixed up: what Google indexes and what AI engines report in their answers. Picture a page with a pricing table, a methodology paragraph and an FAQ section. Google may index the whole document, while ChatGPT, Perplexity, Gemini, Grok or Google AI may surface only one of those sections. Working out which sections emerge is the first step toward reading AI visibility without blending different observations together.
Why Google and AI engines do not read the same way
Google's indexing treats the page as a document, with its sections and fragments. An AI engine's answer, by contrast, is a selected excerpt: it may quote a passage, summarize a block or report a single data point without showing the rest. So a well-indexed page can have sections that are never cited, and a section that barely shows up on Google can emerge in an AI answer. RankAI starts from public pages and names ChatGPT, Perplexity, Gemini, Grok and Google AI as separate observations, not as a single indicator.
An operational method proposed by RankAI
What follows is an operational proposal from RankAI, not a universal standard. Its purpose is to organize the comparison between how pages are read and how AI engines answer, section by section.
- Map the sections of the page. Verifiable criterion: each section has a recognizable heading and standalone content, so it can be cited or ignored on its own.
- Record what Google indexes. Verifiable criterion: compare the indexed page with the section structure, noting which blocks are present and which are not.
- Query AI engines with questions about a specific section. Verifiable criterion: the answer cites that section or reports content from other parts of the page.
- Note the differences between engines. Verifiable criterion: the same section may emerge in ChatGPT but not in Perplexity, or the other way around; the observation must be kept separate for each engine.
The link to the howItWorks page
The reading of pages should be tied to the in-depth page to understand which parts are considered and how the different observations are handled. That is where the method finds its context: knowing that a section is cited is not enough; you need to understand which reading of the page produced that citation.
The most useful point when comparing Google and AI engines is to stop treating the page as a single block. A page can have sections with different jobs: one introduces, one explains a method, one lists criteria, one closes with a limitation. Google may treat them all as parts of the same document; an AI engine may pick only the one that best answers the question asked. That is why RankAI proposes noting three things for each section: whether it is present in the indexed page, whether it was cited in an AI answer, and in which engine. Keeping these three pieces of information separate avoids the most common mistake, which is attributing to one engine a citation that came from a different observation. This is comparison work, not a single measurement: AI visibility reads as a series of differences between sections and between engines, not as one number that sums everything up.
Limits to keep in mind
Claude and Grok are measured only after you connect them. Gemini and Google AI Overviews are not the same observation: read them as two distinct results, even when they concern the same page. An API response is not a person's screen, so what you observe through an API does not match a user's visual experience. These limits should be considered before drawing conclusions about the AI visibility of a section.
In short: AI visibility is measured by comparing sections, not whole pages, and by keeping engines separate. For the full picture of the method, see the in-depth page.