Google AI Overviews visibility, read from the page

Google AI Overviews are the generated answers above classic blue links, with a short list of sources beside or under the paragraph. “AI overview visibility” is the search founders use when their product is missing from that box and a roundup blog is sitting in it. There is no published, stable formula for earning the slot. What you can inspect is whether your page is the kind of source an answer can lift: a direct definition, a named entity, a question with a short reply, and a URL Google can already crawl.

CiteScore’s AI Overviews line is that inspection, priced as the same $39 audit as the ChatGPT and Perplexity lines. We fetch public HTML. We do not run your query through Google.

How the Overview row is decided

The baseline looks at two dimensions more than the others. Structured markup covers JSON-LD types, with extra weight when FAQPage, Question, or SoftwareApplication is present, plus a canonical URL. Answer fitness covers a definitional lead, question-shaped headings, on-page FAQ pairs, and whether the H1 is a slogan. When both are in decent shape, the status can be Mentioned. Otherwise it stays Absent.

Mentioned means “this page has the shape Overviews tend to quote.” It does not mean we saw your brand in an Overview this morning. Absent means the FAQ and entity layer looked thin — the usual state of a homepage that opens on a metaphor and a row of feature tiles. The language-model review, when configured, may nudge the score by a limited amount and rewrite the note. It is told not to claim a live Search result.

On the Northbound sample the Overview line is Absent, with the note that head terms have no inclusion and FAQ / SoftwareApplication markup is missing. That company is fictional. The sentence is the kind of sentence the report writes.

Google-Extended is a different switch

We fetch robots.txt and store a policy for Google-Extended: allow, disallow, or unspecified. Publishers treat that token as “the AI opt-out.” Google’s own description is narrower. Google-Extended controls whether a site’s content is used to improve Gemini Apps and Vertex AI generative APIs. Google says it does not affect inclusion or ranking in Google Search.

So a Disallow of Google-Extended does not, on Google’s published terms, remove you from AI Overviews. It does lower CiteScore’s AI-crawler dimension, alongside GPTBot and PerplexityBot. If you blocked Google-Extended on purpose, treat that part of the crawler score as a record of your policy. The Overview row itself is driven by answer fitness and markup, not by that token. The crawler page lists the three tokens we parse and the ones we skip.

Page changes that match the row

These are the edits the Overview line is built to notice. They are also ordinary good pages. None of them is a trick.

• A definition before the slogan. The first screen should be liftable: what the product is, who it is for, what it replaces.
• Two or three real questions. “What is …?”, “Who is it for?”, “How is it different from …?” with short answers, and FAQPage JSON-LD that matches the visible text. Hidden FAQ markup with no visible FAQ is a bad trade.
• One entity, spelled the same way. Title, H1, and Organization or SoftwareApplication JSON-LD. Overviews cite sources. A source with three names for one product is harder to attach a claim to.
• A date a human would trust. Last reviewed, a changelog, or a current-year article. Freshness is its own dimension. We also note whether /sitemap.xml contains a urlset.

Classic rankings still sit under the Overview. We do not audit backlinks, Core Web Vitals, or index coverage. If the page is noindexed or blocked for Googlebot, fix that in Search Console before you buy an AI-visibility PDF. Our crawler records the generic User-agent: * policy and the three AI tokens. It is not a substitute for a Search audit.

Where this sits next to the other engines

The same PDF has a ChatGPT citation check and a Perplexity line. They share the crawl and diverge on which signals they weigh. ChatGPT is harsher about crawler access. Perplexity is harsher about thin evidence and missing comparison pages. Overviews are harsher about FAQ and entity markup. A founder who only reads the overall number will miss that split. The dimension table is the point of the report.

For the category language — GEO, AEO, “AI visibility” — start with what GEO is or the AEO page. They describe the same $39 deliverable. This page is only the Overview slice.

Questions the Overview line gets