Guide
Query fan-out optimization: how to get cited when AI splits the question
Query fan-out optimization means making your pages the best answer not only to the question a buyer types, but to the several sub-searches an AI engine runs behind it. Google AI Mode, AI Overviews, ChatGPT, Perplexity and Copilot break one prompt into related queries, retrieve results for each, then write one answer. The pages cited are often the ones that rank for those hidden sub-queries, not the head term.
Get a free AEO visibility check
Two lines is enough. A specialist replies within one business day.
The short answer
What is query fan-out optimization?
Query fan-out optimization is the practice of covering the sub-questions an AI engine generates from a single prompt, so your pages rank and get cited for them. In practice: find the fan-out queries for your buyer prompts, group them into a few subtopics, answer each in a clear, extractable section, and track citations over time.
Quotable points
- AI Mode runs far more sub-searches per prompt than ChatGPT or Perplexity, so coverage matters most there.
- Most fan-out queries change between runs; optimize for the subtopic clusters, not single strings.
- Pages ranking for the main query and a fan-out were 161% more likely to be cited in Surfer's research.
- Bing Webmaster Tools grounding queries are the closest first-party view of fan-outs available.
What is query fan-out, and why does it change who gets cited?
Query fan-out is a retrieval technique: the engine reads a prompt, splits it into subtopics, searches for each in parallel and blends the results into one answer. Google described AI Mode in May 2025 as "breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf", and said Deep Search can issue hundreds of searches for one report. Its AI features documentation confirms that both AI Overviews and AI Mode may use the technique.
This is why ranking first for a head term no longer means being cited. In Ahrefs' study of 863,000 keywords and 4 million AI Overview URLs, reported by Search Engine Journal in March 2026, 38% of cited pages ranked in the top 10 for the same query, down from 76% in Ahrefs' July 2025 version. Ahrefs attributed the drop partly to better detection and partly to Google's wider use of fan-out queries.
The practical meaning is simple. An engine answering "best payroll software for a 50-person company" might also search for pricing, compliance, integrations and reviews. If a competitor answers the pricing sub-question clearly and you do not, their page is the one quoted in that part of the answer.
How do fan-out queries differ between AI Mode, ChatGPT, Perplexity and Copilot?
The engines fan out at very different rates, which changes how much breadth each rewards. Surfer's query fan-out research measured an average of 10.7 fan-out queries per prompt in Google AI Mode, about 2 in ChatGPT and 1.4 in Perplexity. The same research found only about 27% of fan-out queries repeated across runs of the same prompt, while roughly 90% fell into four or fewer topic groups.
That last pair of numbers is the key to the whole strategy. Individual fan-out strings are unstable, but the subtopics they cluster into are not. Write for the clusters.
| Engine | Fan-out behavior | What it rewards | Where to see it |
|---|---|---|---|
| Google AI Mode | About 10.7 sub-queries per prompt (Surfer); Deep Search can run hundreds (Google) | Broad topic coverage across several pages | Search Console, Web search type (AI features included, not split out) |
| Google AI Overviews | May use fan-out (Google documentation) | Pages that rank for the main query and at least one sub-query | Search Console, Web search type |
| ChatGPT search | About 2 sub-queries per prompt (Surfer) | A precise page per key sub-question; strong third-party mentions | GA4 referrals from chatgpt.com; manual prompt runs |
| Perplexity | About 1.4 sub-queries per prompt (Surfer) | Tight, well-sourced answers close to the prompt wording | GA4 referrals from perplexity.ai; cited sources panel |
| Copilot | Grounding queries reported by Bing | Pages indexed in Bing that answer the grounding query directly | Bing Webmaster Tools AI Performance |
How do you find the fan-out queries for your buyer prompts?
No engine publishes a full fan-out log, so combine several partial views and treat the result as a working map, not a finished list.
Start with Bing. The AI Performance report in Bing Webmaster Tools shows grounding queries, defined as the key phrases the AI used when retrieving content it cited, with citation share for each. Practical Ecommerce, covering the report's June 2026 improvements, described these grounding queries as the fan-out terms of Bing's AI. It is the only first-party list of fan-outs we know of.
Then run your buyer prompts yourself in AI Mode, ChatGPT and Perplexity, several times each, and record every cited URL. Read the cited pages and note which sub-question each one answers. Add the long question queries in Search Console and People Also Ask. Some tools now generate predicted fan-outs; useful as hypotheses, but check them against what engines actually cite.
What does a fan-out map look like in practice?
Here is an illustrative map for one B2B SaaS buyer prompt, "best payroll software for a 50-person company in Canada". The sub-queries are examples of the kind an engine could plausibly generate, not a captured log, grouped into four clusters, each matched to the page that should own it. Your own runs will produce different wording; the clusters are what to plan around.
| Subtopic cluster | Example fan-out queries | Page that should answer it | Common gap |
|---|---|---|---|
| Cost | payroll software pricing per employee Canada; payroll software cost for small business | Pricing page with plain per-employee figures and a dated table | Price hidden behind a demo form |
| Fit and features | payroll software for 50 employees; payroll with benefits and time tracking | Use-case or segment page answering who it is for | Feature lists with no statement of fit |
| Compliance | CRA payroll remittance software; payroll software that handles provincial tax | Compliance guide with direct answers per requirement | Answer buried in a PDF or support portal |
| Proof and comparison | payroll software reviews; X vs Y payroll | Comparison pages plus current G2 and Capterra profiles | No comparison page; stale review profiles |
How do you optimize content for query fan-out, step by step?
Work cluster by cluster, starting with the prompts closest to revenue. The steps below are the order we use; each depends on the one before.
- 1. Pick 20 to 50 buyer prompts tied to pipeline: category, pricing, comparison and problem prompts, not vanity questions.
- 2. Collect fan-outs from Bing grounding queries, repeated prompt runs, cited URLs and Search Console questions.
- 3. Cluster them into three to five subtopics per prompt, and assign each cluster one owning page.
- 4. Check access first: the owning pages must be indexed in Google and Bing, snippet-eligible and fetchable by OAI-SearchBot, PerplexityBot and Bingbot.
- 5. Rewrite each owning page answer-first: a question H2 per sub-question, a one or two sentence answer, then detail, tables and a visible last-updated date.
- 6. Fill gaps with new pages only where the intent is truly different, such as pricing, alternatives or a compliance guide.
- 7. Close off-site gaps: where engines cite reviews, lists or Reddit threads for a cluster, earn accurate presence there.
How should you structure pages so fan-out queries can extract them?
Engines quote passages, not whole pages, so each section has to stand on its own. Put the sub-question in the heading, answer it in the first sentence, and keep the supporting detail directly below. A pricing section that starts "It depends" loses to one that starts with a number and a condition.
Tables help where the sub-question is a comparison or a set of options, because the facts sit in one place. Name sources in the sentence when you cite data. Keep brand facts identical on your site, in schema markup and on external profiles, so that different sub-queries do not pull different versions of the truth. Schema does not create citations on its own; Google says no special markup is needed for its AI features, so treat it as consistency, not a lever.
Internal links matter more under fan-out than before. When the owning page for each cluster links to the others with descriptive anchors, crawlers and readers can move between sub-questions, and each page carries context for the next.
What mistakes do teams make with query fan-out optimization?
The most expensive mistake is building one page per fan-out string. Because most sub-queries change between runs, those pages chase targets that may not reappear, and mass-produced near-duplicates risk Google's scaled content abuse policy. Surfer's research makes the same point: build topic coverage, not a page per fan-out.
- Optimizing only the head term and assuming the top 10 position carries the citation.
- Treating a fan-out generator's output as fact without checking which pages engines actually cite.
- Answering sub-questions in gated PDFs, support portals or JavaScript-only widgets engines cannot read.
- Ignoring off-site clusters such as reviews and comparisons, where engines often cite third parties.
- Measuring one prompt run and calling it a result; answers vary, so sample repeatedly and read trends.
How do you measure whether fan-out work is paying off?
Measure at the cluster level, monthly. Re-run the same fixed prompt set in each engine and log, for every cluster, whether your page is cited and which competitor is cited instead. Track Bing citation share by grounding query, the Search Console Web search type for the owning pages, and GA4 referrals from chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com.
Expect movement cluster by cluster rather than all at once, and keep a dated change log so you can tie each shift to a specific fix. Our AI visibility tracking service runs this measurement if you would rather not build it in-house.
When is query fan-out optimization not the priority?
If key pages are not indexed, search crawlers are blocked at the CDN, or the site is mid-migration, fix those first. Google's own requirement for AI features is a page that is indexed and eligible for a snippet; fan-out coverage cannot help a page that fails that test. Run an AEO audit or our AEO checklist before investing in content breadth.
AEO Services Agency, the AEO and GEO practice of Greenwolf Tech Labs, builds fan-out maps as part of our AI Overviews optimization work and AEO content writing. Every engagement is fixed-price after a free AEO visibility check, and we measure the same prompt set before and after changes. No one can promise a citation; covering the questions engines actually ask gives you the best odds and the evidence to judge.
FAQ
Frequently asked questions
What is a fan-out query?
A fan-out query is one of the sub-searches an AI engine runs after splitting a user's prompt into subtopics. Google AI Mode and AI Overviews use the technique, as do ChatGPT search, Perplexity and Copilot to varying degrees. The engine blends the results of all sub-searches into one cited answer.
How many fan-out queries does Google AI Mode run?
Surfer's research measured an average of 10.7 fan-out queries per prompt in AI Mode, against about 2 in ChatGPT and 1.4 in Perplexity. Google says its Deep Search mode can issue hundreds of searches for a single report.
Can I see the fan-out queries for my site?
Partly. Bing Webmaster Tools' AI Performance report lists grounding queries, the phrases its AI used when retrieving content it cited, with your citation share. Google does not publish fan-outs, so combine repeated prompt runs, cited URLs and Search Console question queries.
Should I create a page for every fan-out query?
No. Most fan-out strings change between runs, and mass-producing near-duplicate pages risks Google's scaled content abuse policy. Group fan-outs into a few subtopic clusters and give each cluster one strong owning page or section.
Does ranking in the top 10 still matter for AI Overviews?
Yes, but it is not enough. Ahrefs found 38% of AI Overview citations ranked in the top 10 for the query in March 2026, down from 76% in July 2025. Surfer found pages ranking for the main query and at least one fan-out were 161% more likely to be cited.
Is query fan-out optimization different from topic clusters?
It builds on them. Topic clusters organize content around a subject; fan-out optimization starts from the specific sub-questions AI engines run for your buyer prompts, checks which pages get cited for each, and fixes gaps, including off-site sources like reviews.
Explore
Find your industry and market
AEO by industry
AEO by country
Keep reading
Related pages
Get started
Find out what AI says about you
Send your website and three questions your buyers ask. Within one business day we reply with how ChatGPT, Google AI Overviews and Perplexity answer them today, who they cite instead of you, and a fixed-price plan.
- Free AEO visibility check
- Fixed quote before any work starts
- Measured on citations and AI referrals, not promises
Prefer email? info@aeoservicesagency.com
Get a free AEO visibility check
Two lines is enough to start.