When you ask Google's AI Mode a question, it doesn't run one search. It quietly runs many. This process is called query fan-out, and it's the main reason AI search rewards websites that cover a topic in depth across several connected pages.
How query fan-out works
Say someone asks: "What's the best way to take card payments at a market stall in the UK?"
An AI search system might fan that out into searches like:
| Hidden search | What it's looking for |
|---|---|
| card readers for market stalls UK | Product options |
| card reader fees comparison UK | Costs |
| card reader without wifi | Offline and mobile data support |
| tap to pay on iPhone small business | Phone-based alternatives |
| how fast do card reader payments arrive | Payouts |
| card reader for markets reviews | Trust |
It then reads the top results for each, picks the best passages and writes a single answer that covers products, costs, connectivity and payouts.
Google's AI Mode, ChatGPT search and Perplexity all work in broadly this way. As models get smarter, like Google's Gemini 4 Hydrogen upgrade, the fan-out gets more precise and covers more angles.
Why this changes content strategy
In classic SEO, you could target one keyword with one page. With fan-out, a single question touches six or more searches. Two consequences follow:
- More chances to be cited. You don't need to rank first for the main question. Ranking well for one sub-question can earn a citation.
- Depth beats breadth. A site with strong pages on several related sub-questions gets cited more often than a site with one long generic page, and far more than a site with scattered posts on unrelated topics.
What a topic cluster is
A topic cluster is a group of pages about one subject:
- One pillar page answers the main question and gives an overview.
- Several supporting pages each answer a specific sub-question in depth.
- Internal links connect them both ways with descriptive anchor text.
We've written about the classic version in topical authority for SEO. Fan-out makes it more important.
How to build a cluster for AI search
1. Start from a question your customers really ask
Use sales calls, support emails, reviews and Google's "People also ask". Pick a question where you have genuine expertise.
2. Predict the fan-out
List every angle someone needs to decide or act: options, costs, requirements, risks, comparisons, steps, timelines, alternatives. Try the question in AI Mode and note the follow-ups it suggests. Those are fan-out searches in plain sight.
3. Decide what deserves its own page
| Sub-question | Own page? |
|---|---|
| Needs more than 300 words to answer well | Yes |
| Has its own search demand | Yes |
| Is a one-line fact | No, answer it on the pillar page |
| Overlaps heavily with another sub-question | No, merge them |
4. Write each page to answer one thing completely
Lead with the direct answer. Then add detail, examples and a table if it helps. Our AI Mode SEO guide covers what makes a passage citable.
5. Link the cluster properly
- The pillar links to every supporting page.
- Each supporting page links back to the pillar and to closely related siblings.
- Anchor text describes the destination ("card reader fees compared"), never "click here".
6. Add original information to every page
Fan-out gathers many sources for every sub-question. The page with something unique, like your own test, price data or a real example, is the one that earns the citation. Read original research content for ideas that don't need a big budget.
An example cluster
For a bookkeeping firm targeting sole traders:
- Pillar: Making Tax Digital for sole traders: the complete guide
- Supporting pages: MTD thresholds and dates; what a quarterly update contains; software that works with MTD; MTD for landlords; penalties and the first-year grace period; how to keep digital records
Every page answers one real question, links to the others and shares the firm's own experience. That's a cluster that fan-out will find from many directions. (We wrote our own version: Making Tax Digital for Income Tax explained.)
Mistakes to avoid
- Thin supporting pages. Fifty 200-word pages on tiny variations look like scaled content and help nobody.
- Orphan pages. A supporting page nobody links to is hard for Google to find and value.
- Keyword cannibalisation. Two pages answering the same sub-question compete with each other. Merge them.
- Clusters outside your expertise. Depth only helps when it's real.
The short version
AI search splits one question into many hidden searches and cites the best passage for each. Build clusters: a pillar page plus supporting pages that each answer a specific sub-question completely, linked together, with original information on every page. It helps classic rankings and AI citations at once.
Need a content plan built around this? Our content creation team maps clusters from your real customer questions.




