How Google AI Overviews Affect SEO

What is an AI Overview, mechanically?

A composed answer shown above the ordinary results for some queries. Google's own documentation describes the response as grounded: a conventional search runs first, and the answer is drafted from the pages that search returned. So the feature sits on top of the same index and the same retrieval that produce the links below it, rather than running over a separate corpus of its own.

What changes for a page that already ranks well?

The position stays what it was, and the context around it changes. A ranked list with a composed answer above it presents a reader with a stated conclusion before they reach the first link. Whether that changes behavior on any given query is not something this firm can state as a figure. What it changes with certainty is what a report about that query has to record, because a position alone no longer describes what the reader saw.

Falkview's position on the traffic question is that the click-through figures circulating publicly for this feature are not verifiable from outside, and quoting them would be borrowing certainty nobody has. A company's own analytics, read against its own indexed queries over a period, is the only version of that question it can answer for itself.

Does ranking still matter?

Yes, and the documented grounding step is the reason. Where a composed answer is drafted from pages a conventional search retrieved, being retrievable and rankable is the precondition for being drawn from. Work that made a page discoverable and clear still does exactly that, and abandoning it because a new surface appeared would remove the material the new surface reads.

What the feature does undermine is one old assumption: that a strong position implies presence in whatever a search engine says about the subject. Those two can diverge on the same query, and the only way to know is to record both, for the same query, on the same day.

A second assumption is worth retiring alongside it. A page can be cited in the composed answer without appearing high in the list beneath it, and a page can top the list without being cited at all. Neither outcome is a fault in the system, and neither can be read off the other, which is why both belong in the record for the same query.

What kind of page tends to be drawn from?

Speculation about a specific weighting is available everywhere and worth nothing, because no provider publishes one. What can be said from the mechanism as documented is narrower. A page has to be retrievable to be a candidate at all. And a passage is easier to reuse when it holds together on its own, since it arrives in the answer without the heading above it or the page around it.

Falkview expects, and states this as an expectation rather than a finding, that a page answering a specific question directly, in its own words, is easier to draw from than a page circling a topic. The expectation is written down so a later measurement can score it rather than have it confirmed in hindsight.

What should a report change?

Two records for the same query rather than one. The position, with the conditions it was observed under, and separately whether a composed answer appeared, what it said, and which sources it cited. A month-on-month comparison of positions alone can be entirely accurate and still miss the change that mattered most to the company reading it.

Conditions belong in both records. A position varies with location, device and personalization. A composed answer varies between runs and changes when its provider updates the system behind it. A reading quoted without its conditions cannot be compared with the next one, and comparison is the only reason to measure twice.

What should a company actually do?

Keep the technical work, and add the second measurement. Crawlability, indexation, architecture and pages that answer a real question serve both the list and the answer above it. On top of that, a fixed query set checked for both records, so a gap between them is visible while it is still small enough to be worth acting on.

The third piece is the material outside the site. An answer assembled about a company draws on more than the company's own pages, and where independent sources say nothing, the account available is whatever else exists. That work runs at its own pace and is worth starting before a report shows it is needed.

What is not knowable here?

Which queries the feature appears on, and for how long, is decided by Google and changes without notice. How the feature weighs one retrieved page against another is not documented. And how quickly a change on a site reaches a composed answer depends on when the system next revisits the sources involved, which no provider publishes.

A plan depending on any of those three staying still is a plan built on somebody else's roadmap. The durable version is the one that survives all three moving. A site that can be retrieved, pages that state plain facts, independent sources that agree, and a record kept on a schedule so movement is visible when it happens.

Where does this connect?

SEO in the Age of AI Search: What Changed covers what the whole discipline had to absorb rather than what one feature did. What a Search Engine and an Assistant Disagree About, and Why explains what happens when the list below an answer and the answer itself point at different companies. How to Set Up AI Visibility Tracking covers the second record in operational detail.

Falkview scopes the search half as SEO and measures the answer half alongside it on one schedule, in one report, because the two are only useful next to each other.