Your shopping cart is empty!
Search is now a chain of follow-ups, and one keyword page no longer covers it
Publication date: 25.09.2026
In short
- Search is increasingly a chain of follow-ups: what a person asks second and third reveals their goal more precisely than the first question.
- According to Pew Research, after a Google AI overview a user clicks a regular result in only 8% of visits, against 15% without one, and clicks inside the overview itself are rarer still.
- Search Engine Land suggests building topic hubs that hold the reader at every step of their questions, instead of separate pages per keyword.
"What is the best backpack for a five-day business trip?" The opening question is rarely the last one.
Next come the follow-ups: will a laptop fit, will a budget airline allow it, what does it look like in person, what do the alternatives cost, and when will the order arrive.
In conversational search — Google AI Mode, ChatGPT, Perplexity — it is this chain of follow-ups that decides whether a person sees your site at all. Search Engine Land writes about it.
The technique Google calls query fan-out breaks one question into a dozen parallel sub-queries before the person sees an answer. The answer is assembled from fragments of all the sources at once, not from the single page that ranks best.
1%that is the share of visits ending in a click on a link inside a Google AI overview — the rest of the attention stays in the overview.
One query turns into a dozen parallel ones
In the Pew Research study the article cites, after a visit with a Google AI overview a person clicks classic organic results in 8% of cases. Without an overview the figure reaches 15%.
Clicks on the sources inside the overview are rarer still.
For a site that means the fight is no longer over a position for one keyword. The system immediately looks for answers covering clarification, constraints, comparison, validation and action — five kinds of follow-up the authors suggest mapping from real data: search logs, support tickets, sales call recordings, reviews.
One point I'd qualify: Pew measured behaviour in the US, and the figure itself does not say which queries the sample covered. Carrying 8% straight over to Ukrainian results without your own check will not work.
A topic hub instead of a single page per keyword
Instead of one article on "the best backpack for a business trip", the article proposes a system of pages around the topic: a review, a comparison table, a checklist, a case, a short video review, and current data on availability or prices.
The linking logic between such pages should follow the person's chain of questions. It often does not match the site's catalogue structure.
Every claim should rest on a verifiable fact: a method, a test date, the limits of application. An unsupported "best" does not work here.
We covered how to measure a site's presence in AI answers in our guide to measuring AI visibility, while the technical side — indexing, markup, speed — is checked as part of an SEO site audit.
| A page per keyword | A topic hub for a chain of queries |
|---|---|
| Answers one query | Carries the reader through five kinds of follow-up |
| One link in the results | Several entry points for different formats |
| Text without verifiable data | Facts with a method and a test date |
Your own channel that AI cannot reproduce
The authors advise building what an AI system cannot copy from someone else's site: your own benchmarks, tests, calculators, a newsletter or a community. That brings traffic and trust directly, bypassing someone else's summary of the answer.
They suggest measuring four things instead of the familiar top-10 positions: whether you are seen at all across search formats, whether people come back to use the tool or the article again, whether you are cited accurately, and whether any of it reaches a real action — an enquiry, a call, a purchase.
Tellingly, not one of the techniques in the piece is new in itself. Keyword research, internal links and verifiable facts all existed before AI Mode.
What changed is the scale: one page used to cover the whole query, and now it barely covers its first line. I suspect most sites will notice this not in their rankings but in the way traffic to "the main article on the topic" quietly disperses across smaller pages.
