How ChatGPT Fan-Out Queries Actually Search the Web — 15 Million Queries Studied

Type a simple question into ChatGPT — “is [a brand] good for small business accounting?” — and you get back one tidy answer in a few seconds. What you don’t see is the work behind it. ChatGPT usually doesn’t run one search to answer that question. It runs several, quietly, behind the scenes, then stitches the results together before showing you anything.

This invisible step is called ‘query fan-out.’ A new Ahrefs study of 15 million ChatGPT queries reveals exactly how this process changes whether ChatGPT can find and recommend your website.

What “query fan-out” actually means

Here’s the plain version. Say someone asks ChatGPT: “should I switch my business banking to [Bank X]?” A human research assistant handling that question wouldn’t stop at one Google search. They’d probably check a few things — what the bank’s own site says about its accounts, what reviews say, how it compares to competitors, maybe what a forum thread says about actual customer experience.

ChatGPT does something similar. Behind that one question, it can generate several smaller sub-searches: one aimed at the bank’s own website, one aimed at review sites, one aimed at comparison content, and so on. Each sub-search runs independently. ChatGPT then pulls from whichever results look most useful and folds them into the single answer you see. That whole process — one user question breaking into multiple backend searches is what “fan-out” refers to.

It’s worth sitting with that for a second, because it changes what “ranking well in ChatGPT” even means. You’re not competing for one search result. You’re potentially showing up or not across several searches ChatGPT runs without telling you.

The study: 15 million queries, one clear pattern

Ahrefs looked at 15 million ChatGPT queries to understand exactly how this fan-out behavior plays out at scale. Two numbers from that study stand out, and they tell two different parts of the same story.

6.8%: when ChatGPT narrows to one site

Most of the time, a fan-out sub-search is broad — it searches the open web the way a normal search engine would, pulling from whatever sources rank. But sometimes ChatGPT does something more targeted. It uses a site: operator — a search command that restricts results to one specific website to search just that one site instead of the whole web.

According to the study, 6.8% of ChatGPT’s fan-out queries used a site: operator this way. That’s not the majority. Most fan-out searches stay broad. But 6.8% of 15 million queries is still a meaningful, repeatable pattern — ChatGPT isn’t doing this by accident, and it isn’t rare enough to ignore.

75.9%: whose site it goes to

This is the number that actually matters for strategy. Among the most common site:-operator searches ChatGPT ran, 75.9% targeted the brand’s own website — not a retailer selling the product, not a government database, not a forum, not a review site. The brand’s own site, more than any other single category combined.

Think about what that means in practice. When ChatGPT decides to stop browsing broadly and go straight to one source, it’s reaching for the brand’s own pages roughly three times out of four. Retailers, review sites, forums, and government sources are all splitting what’s left.

Why this matters beyond content — it’s about the site itself

Most AI visibility advice focuses on earning mentions in third-party content like listicles, review roundups, and comparison articles. This approach is highly effective. Ridure’s analysis of Ahrefs data reveals that 43.8% of pages cited by ChatGPT are ‘best-of’ listicles, making it the most dominant content format.

However, the fan-out process proves that your own website is also a direct target. When ChatGPT narrows its search, it frequently accesses brand pages directly. Therefore, ensuring your site is clear, current, and easy to crawl is an independent strategy that does not rely on third-party mentions.

 

Where to put GEO/AI-visibility effort

This process does not mean brands should ignore third-party citations; listicle data proves those still carry significant weight. Instead, your own website deserves equal optimization.

Practically, this requires structuring pages clearly and keeping them crawlable so AI systems can easily parse your data. For any effective AI visibility strategy, optimizing your own site is a crucial addition to third-party tracking, not a replacement for it.

FAQs

What is ChatGPT query fan-out?

When ChatGPT answers a question, it often breaks that single question into several smaller searches behind the scenes — the fan-out — runs each one, and combines the results into one answer, rather than running just one search.

How often does ChatGPT search just one specific website?

In Ahrefs’ study of 15 million ChatGPT queries, 6.8% of fan-out queries used a site: operator to search one specific website rather than the open web.

When ChatGPT does search one site directly, whose site is it?

Of the most common site:-operator searches, 75.9% targeted the brand’s own site — more than retailer sites, government sites, forums, or review sites.

Does this mean ChatGPT prefers brand websites over reviews and listicles?

Not exactly. This finding is specific to the smaller set of fan-out queries where ChatGPT narrows to one site with a site: search. Separately, Ridure’s own analysis of Ahrefs data found that 43.8% of pages ChatGPT cites for buyer-style questions are “best-of” listicles, so third-party content still plays a major role in ChatGPT’s answers overall.

What should a brand do with this finding?

Make its own site easy for ChatGPT to search directly — clear page structure, accurate and current information, and crawlable pages — since a brand’s own site is disproportionately the destination when ChatGPT does go direct to one source.

The takeaway

ChatGPT performs multiple invisible searches to answer a single question. While its initial queries remain broad, the model consistently prioritizes a brand’s own pages when focusing on a specific company. This reveals a critical AI visibility strategy that most audits overlook: ensuring your website is optimized for direct AI searches.