Query Fan-Out Explained: Plan Content for ChatGPT Search

Query fan-out is what happens when an AI search system such as ChatGPT takes one question, breaks it into several smaller searches, runs them, and combines what it finds into a single answer. The person types one question; the system quietly asks the web several. So the page that gets cited is often not the one matching the original question word for word, but the one that clearly answers one of those hidden follow-up searches.

This matters now because on 12 October 2026 Ahrefs published a study of 15 million ChatGPT fan-out queries, drawn from 66.6 million free, logged-out ChatGPT responses from August and September 2026. Its findings point to a practical shift: plan your site around the cluster of questions a buyer is really asking, not one keyword per page. Below is a method to do that, with a Jaipur jeweller and a Dubai property firm as worked examples.

What is query fan-out, in plain words?

Think of a good shop assistant. A customer asks which engagement ring to buy. Before answering, the assistant silently checks the budget, the stone type, current prices, what other buyers say and the shop's own policies. Query fan-out is the AI version of that habit. The system rewrites one prompt into narrower searches, often called sub-queries, collects pages for each, and writes an answer from the best passages.

Ahrefs uses the term "fan-out queries" for the searches ChatGPT runs when it goes to the web to answer a prompt. Each one is a separate chance for your page to be found. One broad page competes for one of those searches at best. Clear pages for the follow-up questions can be found several times inside the same answer.

What to check: for your most important service, list five follow-up questions a careful buyer needs answered. If your site answers fewer than half clearly, you have a fan-out gap.

What the Ahrefs ChatGPT query fan-out study found

These findings from the Ahrefs study by Louise Linehan matter most for content planning:

• Not every answer searches. Only 14.4% of the 66.6 million responses contained fan-out queries, with an average of about 1.6 fan-out queries each.
• The current year is the top added term. "2026" appeared in 17.6% of fan-outs in August and 22.9% in September.
• ChatGPT looks for official information and prices. "Official" appeared in 13.1% of English fan-out queries in September, and "price" rose from 6.04% to 8.88% of fan-outs in one month.
• Country context is rising. In the study's table of terms, "india" rose from 3.0% to 6.2% of fan-outs.
• Some searches target one website. 6.8% of the 15 million fan-outs used the site: operator to search a single website.
• Official sources win factual questions. In a separate test of 20,000 US prompts, 84.4% of factual questions about pricing, product features or company details cited at least one official source.

Keep the limits in mind. Ahrefs calls these numbers a snapshot of one model between August and September 2026, and notes that free, logged-out sessions search less than paid or reasoning models. Treat the figures as direction, not fixed rules.

For an Indian or UAE business: the rise of the year, "price", "official" and country names tells you what to state plainly on your own pages: current information, prices or price ranges, and where you operate.

Why query fan-out changes how you plan content

Traditional keyword planning gives each keyword its own page and stops there. Fan-out rewards a group of pages that together cover one buyer's full decision. The Ahrefs guide How to Optimize for AI Search explains that AI search engines can split pages into smaller chunks, retrieve the sections most relevant to a query, and cite only the passages that support an answer. It also argues that for a specific prompt you are often better off with a dedicated piece of content on that exact topic.

The same guide notes that Google says its AI search features still rely on its core search systems. Good pages and a crawlable site still come first; what changes is how you decide which pages to build. For the wider picture, read our guide on AI search optimisation for D2C brands. This post focuses on planning the cluster.

Step 1: Start with one real buyer question

A topic cluster strategy for AI search starts with a question a paying customer actually asks, in their words, not a tool keyword or your internal product name. Find these in sales calls, WhatsApp enquiries, showroom conversations and reviews.

• Jaipur jeweller: "Is a lab-grown diamond engagement ring a good choice for my budget?"
• Dubai property firm: "Should I buy an off-plan apartment in Dubai as an overseas investor?"

Pick a question that leads to revenue and that you can answer with real expertise. One well-mapped question beats twenty loosely chosen keywords.

What to check: confirm with your sales team that they hear this question at least weekly.

Step 2: Fan the question out into a question map

Do the fan-out yourself before the AI does. Write the main question at the top of a sheet and list the follow-up searches a careful system might run, sorted into these angles:

1. Meaning: what is it, how does it work.
2. Price: what does it cost this year, what changes the price.
3. Comparison: this versus the alternative.
4. Official facts: specifications, certification, policies.
5. Proof: reviews, and who it suits or does not suit.
6. Place: the city or country context.
7. Process: how to buy and what happens next.

For the Jaipur jeweller, the map might include lab-grown vs natural differences, lab-grown ring prices in India in 2026, how certification works, the exchange policy, and buying in Jaipur vs online. A useful shortcut is to type the main question into ChatGPT and Google and note which angles their answers cover.

This is where keyword research for AI search changes. Keyword tools still help you choose which main questions deserve a cluster, but many sub-questions show little or no volume in tools, and the Ahrefs study does not report search volume for fan-out queries. Do not drop a sub-question just because a tool shows zero.

What to check: a good map has roughly 8 to 20 sub-questions. Far more usually means you have drifted into a different topic.

Step 3: Group the sub-questions into pages

Group sub-questions by intent, meaning what the person is trying to get done. Questions one focused page can answer belong together; questions that need different evidence or serve different decision stages get separate pages.

The Jaipur jeweller's cluster could be:

• Hub: Lab-grown diamond engagement rings: an honest buyer's guide.
• Lab-grown vs natural diamonds: what actually differs.
• Lab-grown diamond ring prices in 2026, and what changes them.
• How to read a diamond certificate.
• Our exchange and buyback policy.

The Dubai property firm's cluster could be:

• Hub: Buying off-plan property in Dubai as an overseas investor.
• Off-plan vs ready property: how risk and payment differ.
• Payment plans explained, project by project.
• The buying process from booking to handover.
• Service charges and running costs to budget for.

Link every supporting page to the hub and the hub to each supporting page. Our post on content marketing for Indian D2C brands covers keeping a publishing rhythm once the map is set.

What to check: no two pages should answer the same sub-question. If they do, merge them.

Step 4: Write each page so a passage can stand alone

Because AI systems retrieve chunks, each section should make sense on its own. The Ahrefs guide says AI systems do not always process a page from top to bottom, so key facts should make sense even when pulled out of context. It also cites research by Kevin Indig finding that 44.2% of AI citations came from the first 30% of a page.

• Put the direct answer in the first two or three sentences of the page and of each section.
• Use headings that repeat the sub-question in plain words.
• Name the subject in each section instead of writing "it", so a lifted passage still reads correctly.
• Keep key facts in visible text. In an Ahrefs test, AI systems ignored JSON-LD and hidden markup and relied on visible HTML.

For page-level checks, see our checklist for getting cited in Google AI Mode.

What to check: paste any single section into a blank document. If a stranger understands it without the rest of the page, it is ready.

Step 5: Publish the facts only you can publish

The Ahrefs study's advice is direct: "The pages holding up best are the ones only you can publish." It recommends that your own site states factual details such as prices, specifications and support information in plain text on a page ChatGPT can find, which is what those site: searches need.

For the Jaipur jeweller, that means pages explaining making charges, certification and the exchange policy. For the Dubai property firm, it means current payment plans, handover timelines and service charge information for each project. Use real, dated figures and update them when they change. Jewellery brands will find more ideas in our jewellery brand marketing playbook for India.

What to check: if your prices and policies live only in PDFs, images or chat replies, move them onto plain web pages.

Step 6: Add the comparison and review pages fan-outs look for

Product pages alone rarely cover a full decision. The Ahrefs study puts it this way: "A running shoe brand would get more from review coverage and comparison pages than from its product pages alone." The same logic applies to most businesses.

• Write honest comparison pages, including when the alternative is the better fit.
• Publish real reviews, with permission, on pages that explain who the product suits.
• Earn coverage on credible third-party sites in your sector.

For UAE brands, our post on content strategy for luxury retail in the UAE covers trust signals for these pages.

What to check: list the top three alternatives buyers consider. You should have a fair comparison page for each.

How to check whether it is working

Set a baseline before you publish and review monthly.

• Search Console Performance report: filter by each cluster page and note impressions, clicks and the queries it appears for. Longer, question-style queries appearing for supporting pages are an early good sign.
• Hub growth: compare the hub page's impressions and clicks for the three months after launch against the three months before.
• Manual AI checks: monthly, ask ChatGPT your main question and three sub-questions in a logged-out session, and record which sites it cites in a simple sheet.
• Enquiry quality: ask your sales team whether new leads mention topics from the cluster pages.

If an agency reports these numbers to you, our guide to reading your marketing agency's reports explains what to ask.

Frequently asked questions

What is query fan-out in SEO?

Query fan-out is when an AI search system splits one question into several related searches, gathers pages for each and combines them into one answer. In SEO, it means planning pages for the follow-up questions a buyer asks, not only the main keyword. Each of those searches is another chance to be cited.

Does ChatGPT always use query fan-out?

No. In the Ahrefs study of free, logged-out ChatGPT responses, only 14.4% of responses contained fan-out queries, with about 1.6 each on average. Ahrefs notes that free sessions search less than paid or reasoning models, so this varies.

How many pages should a topic cluster have?

There is no fixed number. A cluster built from one buyer question often ends up with a hub page and four to eight supporting pages. The rule is that each page answers a distinct sub-question.

Is keyword research still useful for AI search?

Yes. Keyword tools show which main questions are worth building around and help you track progress in normal search. Use them to choose the hub, then use your question map and customer conversations to choose supporting pages.

If you would like help turning your best buyer questions into a question map and a cluster of pages that both Google and ChatGPT can use, our SEO team can build the map with you and write the pages. For businesses in Rajasthan, start with our SEO services in Jaipur.