What is query fan-out, and why does it matter for GEO?

A homeowner with a steep TNB bill asks ChatGPT for a solar installer for her terrace house. She reads one tidy answer. Behind it, ChatGPT ran several web searches she never typed. That's query fan-out: the AI turns one question into searches of its own, usually several, and answers from what those searches find.

For GEO, that changes which words your pages have to answer. In our test of Malaysian buying questions, only 1 of 401 searches repeated the customer's question word for word.

What is query fan-out in ChatGPT and Google's AI?

Query fan-out is the step where an AI assistant rewrites one question into one or more targeted web searches, often called sub-queries, before answering. OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries"¹.

Google uses the name itself. Its documentation says "Both AI Overviews and AI Mode may use a 'query fan-out' technique"², issuing related searches across subtopics. Google's Head of Search, Elizabeth Reid, describes AI Mode "breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf"³.

What does query fan-out look like for a Malaysian customer's question?

Query fan-out on a Malaysian question looks like ChatGPT quietly adding the local details a customer would check next. We asked ChatGPT: "solar panel installer for a terrace house in Malaysia to cut my TNB bill". In one run, it wrote these three searches:

• "Malaysia residential solar panel installer terrace house TNB bill net energy metering 2026 site:.my"
• "Malaysia rooftop solar installer residential terrace house TNB bill solar quota 2026 site:.my"
• "TNB solar for home Malaysia rooftop installer residential 2026 site:tnb.com.my OR site:seda.gov.my"

The customer never mentioned net energy metering (NEM), quotas, SEDA or the year. Our reading is that ChatGPT added them because a careful buyer checks them next. In two other runs it also searched for a "quote Kuala Lumpur" and for "financing".

So one question became several sub-questions: which scheme applies, is there a quota, what do TNB and SEDA say, what does it cost, can I get financing. An installer's page that answers those has more ways to be found.

What does ChatGPT add to its searches when it fans out a Malaysian question?

ChatGPT's fan-out searches for Malaysian buying questions added the same kinds of detail again and again. It searched in 130 of our 135 answers (CSV). A typical answer came from three searches.

Peec AI's analysis of millions of fan-out searches found "ChatGPT: 2.1 fanouts per query"⁴. Our reading is that the question, model and setup all change the count.

Regulators showed up most for solar, in 14 of 24 solar answers. Every search is in the CSV.

Language matters too. Asked in Malay or Chinese, ChatGPT still wrote most of its searches in English, as we showed in a separate test.

Why does query fan-out matter for GEO and my business?

Query fan-out matters for GEO (generative engine optimisation: getting AI tools to name your business) because ChatGPT's own searches, not your customer's exact words, decide which pages it can draw on. Our reading is that a page stating the details ChatGPT searches for gives it more chances to match. In our test, those were full place names, reviews, regulators and registrations, and price words when the customer asked about cost.

Don't chase the exact search strings, though. They change: only 4% of the searches in one run reappeared word for word in another run of the same question (data). Write for the sub-questions instead. In the solar example, net energy metering came up in all three runs, though no two searches matched.

And don't build a page for every search. Google says "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary"². Our reading: one complete, specific page serves fan-out better than dozens of thin ones.

What can't our test tell me about query fan-out?

Our test shows the searches ChatGPT wrote, not everything that happens after them:

• Which search mattered. We saw the searches and the answer, not which search led to which business being named.
• The ChatGPT app and Google. We tested OpenAI's API with default settings on one evening. The app, Google's AI Mode and other tools may fan out differently.
• Every trade. The 45 questions covered five trades, so a clinic or café may see different added terms.

How do I optimise my business pages for ChatGPT's query fan-out?

Most customer questions in our test reached ChatGPT's search as several smaller ones. Each is one more chance for your page to be found.

1. Request EasyFound's free check. Find out what ChatGPT tells your customers about you, then talk it through with us on a free call.
2. List the sub-questions behind each customer question. Think price, area, licence, warranty and reviews. A tool that predicts fan-out gives you a guess. Our CSV shows what ChatGPT actually wrote in one test, across five trades.
3. Answer them on your own pages in plain text: the areas you serve with full names (Kuala Lumpur, Petaling Jaya) and your registration or licence numbers. Add what customers check next too: your price range in RM, warranty terms and the languages you serve in.
4. Keep the page current. In our test, ChatGPT added the year to most answers' searches, far more often than Peec AI saw in its data. Our advice: show when your prices and details were last updated.

EasyFound is an AI SEO service for Malaysian businesses. We check what ChatGPT and five other AI tools say in English, Malay and Chinese, then fill your pages with the answers their searches look for. Start with the free check below.