Guide
Why does Google's own AI Overview skip the page that ranks first?
Because the AI Overview box is built from a different, wider search than the one that produced the ranked list underneath it. Google runs your query again behind the scenes, breaks it into several smaller questions, and pulls in whichever pages answer each one best. Ranking first in the organic results answers the question you typed. It says nothing about how well your page answers the four or five sub-questions Google generated from it, and those are what the box is actually built from.
This is a recent and fairly sharp change, not a permanent feature of how Google has always worked. Ahrefs studied 863,000 keywords and 4 million AI Overview citations and found that in mid-2025, 76% of cited pages also ranked in the top 10 organically. In its most recent pass, that had fallen to 38%. The remaining citations now split almost evenly between pages ranked 11 to 100 and pages ranked beyond 100 (Search Engine Journal, 2026). A page ranking first is now roughly as likely to sit outside the AI Overview’s sources as inside them.
Why ranking first stopped being enough
Take a business that sells commercial espresso machines and ranks first for “best commercial espresso machine.” That query looks simple, but a shopper typing it is really asking several things at once: which machines are reliable, what they cost to run, which brands technicians can actually service, and which one suits a busy café rather than a quiet office. Google’s synthesis pass treats the query the same way. It generates something close to that list of sub-questions, searches each one separately, and assembles the box from whichever pages answered each part convincingly.
The page ranking first for the main query might cover reliability well and say nothing about running costs or servicing. A page ranking eighth might be a forum thread comparing running costs across five machines line by line, or a manufacturer’s own spec page listing service network coverage by region. Neither of those would beat the top page on the original ranking signals: backlinks, domain authority, on-page relevance to the exact phrase. Both can out-answer it on one sub-question, and that is enough to get pulled into the box while the top-ranked page is left out of it.
This is also why the box can cite a page that ranks nowhere in the top 100. Ahrefs found 31% of AI Overview citations came from outside the top 100 organic results entirely. A page can be a poor general match for the query and still be the best specific answer to one fragment of it.
What this means in practice
It means treating a page’s ranking position as a proxy for AI Overview visibility is now wrong more often than it is right. A marketer who optimises a page to climb from third to first on the organic list, using the signals that used to move rankings, may see no change at all in whether that page gets cited above the fold. The lever that used to work is answering a question the synthesis pass is no longer asking.
The practical shift is to write for the sub-question rather than the headline query: a section that states running costs in numbers, a paragraph naming which service networks cover which regions, a comparison a reader could act on without leaving the page. That is closer to what a forum answer or a spec sheet does by default, which is part of why pages like that show up in the 62% Ahrefs found outside the traditional top 10.
More on this shift is in Have People Stopped Googling and Started Asking ChatGPT Instead?
Common questions
Does this mean SEO no longer matters? No. Ranking well still drives the organic clicks below the box, and a page still has to be found and judged relevant before it can be pulled into any sub-question search. It means ranking first is no longer sufficient on its own for the box above those results.
Will optimising harder for the main keyword fix this? Not by itself. The sub-questions a synthesis pass generates from a query are not visible in any keyword tool, so the fix is answering the specific things a buyer would actually ask next, not chasing the same phrase harder.
Does every query get an AI Overview? No, and Google does not publish which ones trigger it. Queries with a clear informational or comparison intent are the ones most likely to produce one.
Courage Horizon sets up AI agents for UK small businesses: software that does one job, such as chasing unpaid invoices, and asks you before it sends anything. Being found when customers ask ChatGPT is part of that work. Get a free report on your own business.