Query fan-out is when an AI assistant turns one question into several narrower web searches, runs them, and builds its answer from what comes back. Google uses the term publicly to describe how AI Mode works, and similar behavior can be observed in ChatGPT and other assistants when they browse. The pages that rank for those sub-queries are the pool the assistant cites from.
In other words, the assistant does not always search for exactly what you typed. Those hidden searches decide which pages it sees, and therefore which brands it can recommend. Understanding them changes how you plan content for AI visibility.
What query fan-out looks like
Take a prompt like “What is the best CRM for a small B2B startup?” An assistant that decides to search might run queries along the lines of:
- best CRM for startups 2026
- CRM for small B2B sales teams reviews
- HubSpot vs Pipedrive for startups
- affordable CRM with free plan
These examples are illustrative, but the pattern is common: the assistant expands the question into angles it thinks will help, such as recency, segment, comparisons and price. It then reads some of the results and writes an answer, often citing a few of the pages it found.
The prompt you track and the searches that actually happen are different things. That gap is where useful insight lives.
Why query fan-out matters for your visibility
The assistant can only cite what it finds
If the assistant searches “CRM for small B2B sales teams reviews” and your brand does not appear in those results, it is unlikely to show up in the answer, however good your homepage is. Fan-out queries define the candidate set.
Sub-queries reveal what the assistant thinks matters
When an assistant adds “free plan” or “integrates with Gmail” to its searches, it is signaling the criteria it is using to judge options. Those are the attributes your content should address clearly.
It connects GEO back to SEO
Fan-out queries are ordinary web searches. You can research them, check who ranks, and create or improve content to compete. It is one of the most direct bridges between classic SEO and generative engine optimization. For the bigger picture, see what is generative engine optimization.
How to see ChatGPT fan-out queries with Citegram
Visibility into these searches depends on the assistant. Some show the searches they ran in the interface, some show only the sources, and some show neither. What is exposed can also change as products evolve.
Citegram, a Chrome extension, records the web searches ChatGPT ran for each tracked prompt. The workflow:
- Add your priority prompts. Start with the buyer questions you already track. Our guide on choosing prompts to track for AI visibility helps you pick them.
- Run them in the real app. The extension asks each prompt in ChatGPT in a background tab, using the account already signed in to your browser, so no API key is needed.
- Read the fan-out queries. For each answer you see the searches ChatGPT ran, alongside the answer text, whether and where your brand is mentioned, and the cited sources in rank order.
- Repeat the runs. Fan-out changes between runs, so re-running the same prompts builds a fuller picture of which searches recur.
- Compare with other assistants. The same prompts can run in Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, so you can see whether the sources found through fan-out show up elsewhere too.
If you do it manually, run each prompt in a fresh chat, note any searches displayed, and repeat a few times, since the queries change from run to run.
A process for using fan-out data
- Collect. For each tracked prompt, gather the fan-out queries across several runs.
- Group. Cluster similar queries. “best CRM startups 2026” and “top startup CRM this year” belong together.
- Count recurrence. Queries that appear in most runs matter more than one-offs.
- Search them yourself. For each recurring query, check which pages rank in regular search.
- Map your coverage. Do you have a page that answers this query directly? Are you mentioned on the pages that rank?
- Close gaps on your site. Create or update pages that match recurring sub-queries, especially comparisons, use-case pages and pricing explainers.
- Close gaps off your site. Where third-party pages rank, work on being included honestly through reviews, accurate briefings to publishers and fair comparisons. See how to get cited by AI.
- Re-check. Keep tracking. Watch whether your pages start appearing in cited sources for the prompt.
Common query fan-out patterns worth watching
- Year modifiers. Assistants often add the current year to look for recent content. Keep key pages updated and dated.
- “Reviews” and “reddit”. Signals that independent opinion is being sought. Your review profiles and community presence matter here.
- “vs” queries. Even when the prompt is generic, the assistant may search head-to-head comparisons. Fair comparison pages help.
- Constraint terms. Price, team size, integrations and industry show up as qualifiers. Make these explicit on your pages.
- Brand-specific checks. Sometimes the assistant searches a specific brand name to verify details. Make sure your own pages give clear, current facts.
Fan-out analysis checklist
- Fan-out queries collected for every priority prompt
- Queries grouped into clusters
- Recurring queries identified across runs
- Ranking pages checked for each recurring query
- Coverage gaps listed for your own site
- Coverage gaps listed for third-party sources
- Content updates assigned with owners and dates
- Tracking in place to measure citation changes
Limits to keep in mind
Fan-out is not deterministic. The same prompt can produce different searches on different days, and assistants do not publish exactly how they decide what to search or which results to read. Use the data to find patterns and priorities, not as a precise recipe. Also remember that not every answer involves a web search. Some answers come from what the model already knows, in which case there is no fan-out to observe.
FAQ
What is query fan-out in simple terms?
Query fan-out is an assistant breaking your question into several smaller web searches before it answers. It reads results from those searches and combines them into one response. That is why the pages it cites can differ from the top results for your original question.
Is query fan-out the same in every assistant?
No. The general idea appears in multiple assistants that browse, but how often they search, how many queries they run and how much they show users differ. Treat each assistant separately.
Should I write a page for every fan-out query?
No. Many sub-queries overlap, and a single strong page can cover a cluster. Focus on recurring queries that match your buyers and where you have something useful to say.
Do fan-out queries replace keyword research?
They complement it. Keyword tools tell you what people search. Fan-out queries tell you what assistants search on their behalf. The overlap between the two is often a good place to start.
See the searches behind your answers
If you want to know which searches ChatGPT runs before answering your most important question, the free plan lets you track 1 prompt and see the fan-out queries, cited sources and answers over time. Pro adds unlimited prompts. Take a look at the pricing page.
ChatGPT
