Search is splitting into two channels — traditional rankings and AI-generated answers. Here's what we track differently for each, and why most "AI SEO" advice misses the point.
The two channels are not the same game
Traditional search still works the way it always has: a query, ten blue links, and a click. Position and click-through rate are the currency. AI search — AI Overviews, ChatGPT, Perplexity, Gemini — works differently. Instead of a ranked list, the system synthesises a single answer, usually citing two to four sources. There is no position five. Either your business is part of the synthesised answer, or it isn't.
That binary nature is why treating AI visibility as "just SEO with an extra checkbox" undersells it. The retrieval mechanics are different: these systems favour content that answers a specific question clearly and early, over content that is merely comprehensive or well-linked. A page that ranks position three in Google is not automatically the page an AI model chooses to cite.
What we actually check
Extractability — whether a page answers the likely question directly, within the first two or three sentences after a heading, rather than building up to it after three paragraphs of scene-setting.
Structured data — schema.org markup (FAQPage, Article, Organization, Product) makes content machine-parseable. It doesn't guarantee a citation, but it removes ambiguity that otherwise works against you.
Consistency and freshness — AI systems weight sources that are internally consistent and current. Contradictory facts about the same business across different pages (old addresses, outdated pricing, mismatched descriptions) actively hurt this.
Off-site corroboration — mentions and citations elsewhere on the web function similarly to how backlinks work for traditional SEO: they're a trust signal that this business is who it says it is.
How we monitor it
We run a fixed set of representative, buyer-intent queries through the major AI search surfaces on a recurring basis and record whether — and how — the brand appears. This is deliberately qualitative rather than a single rank number, because AI-generated answers vary between runs even for the same query. The useful signal is the trend across weeks, not any single response.
What this means practically
Technical SEO is still the foundation — a page that isn't indexed can't be cited by anything. What changes is the writing discipline: write to directly answer the question a real buyer would type or speak, not just to rank for a keyword phrase. Structured data stops being a nice-to-have. And measurement has to include a channel that didn't exist three years ago, which is exactly why our audits report on traditional ranking factors and AI-citation factors as two related but separate tracks.