The bottom line:
- ChatGPT doesn’t run one search. It runs several. Your prompt gets decomposed into a sequence of queries, and that sequence is the fan-out.
- The first query is the prize. It’s the primary citation source in 67% of multi-query prompts. Ranking #3 on the first query beats ranking #1 on the last one.
- You only need to win a few queries. Queries 1-3 account for 92% of the citation coverage a domain can get from a fan-out.
- SERP position matters less the deeper the fan-out goes. The gap between #1 and #10 narrows on later queries, so classic rank-chasing pays off least exactly where most content competes.
What query fan-out actually is
Ask ChatGPT something that needs current information and watch what it does. It doesn’t answer from memory and it doesn’t run your prompt as a single search. It generates a list of search queries, runs them in order, reads what comes back, and synthesizes an answer with citations.
That generated list is the query fan-out. Three variables then decide whether your page gets cited:
- The fan-out itself: which queries the model generates for a prompt.
- Query position: where in that sequence a query sits.
- SERP position: where your page ranks on Google for that query.
Traditional SEO only ever had the third lever. AI search added the first two, and they turn out to dominate.
This isn’t a ChatGPT quirk. Google describes AI Mode with the same term, query fan-out: the model issues multiple related searches on your behalf and assembles the answer from the union of results.
What the data showed
The best public measurement of this came from a 2025 study by Profound, which compared ChatGPT’s citations against Google results across 420 prompts and 2,867 generated queries. The original write-up has since been taken offline, but the findings are preserved in the Internet Archive, and they’re worth restating because the numbers change how you should think about visibility.
The first query dominates
The first query ChatGPT generates drove 34.4% average domain overlap with citations, nearly double the 17.6% of every query after it. In 67% of multi-query prompts, the first query contributed more citations than any other position.
Position and query order multiply
A #1 Google ranking on the first fan-out query produced a 40.2% domain citation rate. The identical #1 ranking on the last query produced 24.3%. Same page, same rank, 1.7x difference purely from where in the fan-out the query sat.
The starkest version: ranking #3 on the first query (26.0%) beat ranking #1 on the last query (24.3%). And first queries punish slippage hardest: dropping from #1 to #6 there cost 23.5 percentage points of citation rate.
Deeper fan-outs dilute every query
Low-fan-out prompts showed 32.4% domain overlap per query. High-fan-out prompts dropped to 12.7% per query. The more searches the model runs, the less any single one decides the answer.
Late queries flatten the SERP
The study called it gradient compression: on later queries, the citation-rate gap between ranking #1 and ranking #10 narrows. The model gets less picky about rank as it fills in details. Which means the place where a #1 ranking is genuinely worth fighting for is early in the fan-out, not late.
A few queries carry almost everything
Cumulative domain overlap reached 45.6% by the fifth query, but 92% of that maximum came from queries 1-3. Winning three queries is effectively winning the fan-out.
What to do with this
Target the first query, not the long tail. First queries are the obvious, direct phrasings: “project management software”, not “project management software for remote teams with time tracking”. A decade of long-tail SEO advice inverts here. The specific phrasing you’d normally chase tends to appear later in the fan-out, exactly where citation rates are lowest.
Watch the fan-out for your category. ChatGPT shows the searches it runs while it works. Prompt it the way your customers would, note the queries it generates first, and treat that list as your keyword research. The fan-out is observable; most competitors never look.
Aim for top 3 on the queries that matter. You don’t need #1 everywhere. You need top 3 on the two or three queries that open the fan-outs in your category, because that’s where position sensitivity is steepest.
Match one page to one first query. URL-level overlap on first queries ran 18%, roughly 3x the rate of middle queries. The model cites pages whose content aligns directly with the query it just ran, so a focused page beats a sprawling one.
Stop over-indexing on deep-tail rankings. A #1 ranking that only surfaces on query six of a fan-out is worth a fraction of a #3 on query one. Budget accordingly.
Zusammenfassung auf Deutsch
Wenn ChatGPT eine Frage beantwortet, führt es mehrere Suchanfragen aus, nicht eine. Die Analyse von Profound (420 Prompts, 2.867 Suchanfragen, Original nicht mehr online, archiviert) zeigt: Die erste Suchanfrage im Fan-out liefert in 67% der Fälle die meisten Zitate. Platz 1 bei der ersten Anfrage bedeutet eine Zitierrate von 40,2%, dieselbe Platzierung bei der letzten Anfrage nur 24,3%. Platz 3 bei der ersten Anfrage (26,0%) schlägt Platz 1 bei der letzten (24,3%). 92% der maximalen Abdeckung stammen aus den Anfragen 1-3. Praktische Folge: Optimieren Sie auf die naheliegendste, direkteste Formulierung eines Themas, nicht auf Long-Tail-Varianten, denn diese erscheinen früh im Fan-out, wo Rankings am meisten zählen.
Where this leaves you
Query fan-out is the clearest mechanical difference between ranking on Google and getting cited by AI, and it’s measurable today. If you want to know which queries open the fan-outs in your category, and whether your pages rank where it counts, that’s exactly what Fokal tracks for your brand across Google and AI search.