How does Google decide what goes into the AI answer?
Definition
According to Google's own AI Mode blog post, the query fan-out technique fires multiple related searches at the same time, across different subtopics and sources, and gathers the results into one generated answer. That is different from classic ranking, where one page competes for one position against every other page for the same query.
One question becomes several searches
Imagine a buyer asking an AI assistant which type of valve resists a specific chemical at high temperature. Instead of matching that sentence to one ranked page, the system can split it into separate searches: what the chemical is, which materials resist it, which valve types exist for that application, and which suppliers publish that combination. Each of those sub-searches has its own set of results.
From One Question to Several Answers
You no longer compete for one position
On a page of ten blue links, a page below position ten was invisible. Inside a fan-out answer, a page can be invisible for the exact phrase the buyer typed and still be picked as the best fragment for one of the sub-questions the system generated. That is a different game, and it rewards a different way of writing.
For Roberto Wydra, founder of RudekWydra, the practical consequence for an industrial company is that a single long page trying to cover an entire product line loses to several shorter pages, each one answering one sub-question with a clear, self-contained paragraph. "The fan-out rewards the fragment that answers by itself, without needing the rest of the page to make sense", he says.
Write the sub-question before writing the section
Three Habits That Fit the Mechanism
The structured way to make this legible to a system, not only to a person, is described in how to structure content so an LLM cites it.
Writing to rank, not to be picked
Most industrial pages are still written to satisfy one keyword and one crawler pass: the term repeated in the title, in the first paragraph, in the alt text. That habit ranks a page for the exact phrase and does nothing for the sub-questions a fan-out generates around it. The result is a page that shows up in classic search and never gets pulled into a generated answer.
The fix is not writing more — it is writing in smaller, answerable units, each one addressing a question the fan-out is likely to generate. That is a mechanism observable in how the system works, not a measured ranking gain, and it is the reasoning behind this hub’s emphasis on organizing technical content before buying more media.
FAQ
How does Google decide what goes into the AI answer? +
What is query fan-out? +
Does my page need to rank first to appear in an AI answer? +
What should a page look like to be picked as a fragment? +
Is this the same as classic SEO ranking? +
Does a long page that covers everything perform better in a fan-out answer? +
Where can I see this idea applied to writing for an LLM? +
Does fan-out explain why my organic traffic fell even with good rankings? +
Find out if your content survives being one fragment
The diagnosis checks whether your technical pages answer questions on their own, or only make sense as part of a longer page nobody generates a fragment from.