Query Fan-Out: How to See the Exact Queries AI Runs Behind Your Keyword (Free Methods Included)

Type one question into Google's AI Mode and Google may run a dozen searches on your behalf. You optimized for the question; the citations went to whoever answered the twelve searches.

That mechanism is called query fan-out, and it explains most cases of "we rank #1 but ChatGPT never cites us." This guide covers how it works, and then goes where the existing coverage doesn't: how to extract the actual fan-out queries yourself, for free, and what to do with them.

How Query Fan-Out Works

Fan-out follows a four-step pipeline, and each step is a distinct place where your content wins or loses.

Step 1: Decomposition. The model reads the prompt, judges whether it's complex enough to need multiple angles, and generates sub-queries covering the facets it predicts the user cares about. A prompt about "best CRM for a small agency" spawns sub-queries about pricing, comparisons, integrations, and reviews.

Step 2: Parallel retrieval. The system runs every sub-query at once against its index or the live web. This is the structural break from classic search: one input, many simultaneous retrievals.

Step 3: Evaluation and extraction. For each sub-query, the model scans top results and extracts the specific passage that answers it. It doesn't read your page the way a person does; it lifts chunks.

Step 4: Synthesis. The extracted chunks get woven into one response with citations. A single answer routinely cites one site for pricing, another for reviews, and a third for specs, sources that would have competed for one ranking slot in classic search.

Losing at step 1 means the model never generated a sub-query your content answers. Losing at step 3 means it did, retrieved your page, and couldn't cleanly extract the answer. Those are different problems with different fixes, and the rest of this guide treats them separately.

How to Extract Real Fan-Out Queries for Free

Every article on this topic tells you fan-out queries exist. Almost none show you how to pull them. Four methods, ranked by usefulness.

Method 1: Pull Fan-Out Queries From the Gemini API

{
  "candidates": [{
    "groundingMetadata": {
      "webSearchQueries": [
        "best CRM for small nonprofits 2026",
        "nonprofit CRM under 50 dollars per month",
        "Salesforce Nonprofit Cloud vs Bloomerang",
        "CRM with QuickBooks integration for 501(c)(3)"
      ],
      "groundingChunks": [
        { "web": {
            "uri": "https://example.com/nonprofit-crm-guide",
            "title": "Best CRM for Small Nonprofits (2026)"
          } },
        { "web": {
            "uri": "https://example.com/bloomerang-vs-salesforce",
            "title": "Bloomerang vs Salesforce Nonprofit Cloud"
          } }
      ]
    }
  }]
}

The Gemini API returns the actual search queries it generated inside the grounding metadata of any grounded response. This is the closest thing to reading Google's homework, and it costs fractions of a cent per prompt.

from google import genai
from google.genai import types

client = genai.Client(api_key="YOUR_KEY")

response = client.models.generate_content(
    model="gemini-2.5-flash",  # use the current grounding-capable model
    contents="best crm for a small marketing agency",
    config=types.GenerateContentConfig(
        tools=[types.Tool(google_search=types.GoogleSearch())]
    ),
)

meta = response.candidates[0].grounding_metadata
print(meta.web_search_queries)   # the fan-out queries
for chunk in meta.grounding_chunks:
    print(chunk.web.uri, chunk.web.title)   # who got cited

Two outputs matter here. web_search_queries is the fan-out list, and grounding_chunks shows which URLs won each retrieval, which turns this into a citation-competitor report as well.

Run the same prompt five times and diff the lists, then run your ten most valuable prompts through a loop and export to CSV. That's a fan-out research workflow for the price of a coffee, no enterprise platform involved.

Method 2: Watch ChatGPT's Searches

ChatGPT exposes its search activity in the interface: expand the search step on any response and the queries are listed. The free Keyword Surfer Chrome extension also surfaces them alongside the chat.

This method costs nothing and reflects real behavior, but it's manual. Use it to spot-check your highest-value prompts rather than for bulk research.

Method 3: Read AI Mode's Disclosure

Google AI Mode sometimes announces its work, showing that it's kicking off a specific number of searches for a complex query. You won't always get the full query list, but the count alone tells you how aggressively a topic fans out.

Complex comparative prompts triggering many searches signal high fan-out surface area, meaning more sub-queries you can win. A prompt that triggers no visible search activity is being answered from model memory, and that's a different visibility problem entirely.

Method 4: Ask the Model (with a Caveat)

Prompting "what searches would you run to answer X" produces a plausible facet list, and it's useful for brainstorming. It is not a log of real behavior, so treat the output as hypothesis, never as data.

What Fan-Out Queries Actually Look Like

Fan-out queries cluster into predictable facet types regardless of topic. Pull the queries for any commercial prompt and you'll see the same skeleton.

The individual query strings change between runs. The facet skeleton barely moves, and that distinction drives the entire optimization strategy below.

Query Fan-Out Providers & Pricing

Developers building their own Answer Engine Optimization (AEO) apps rarely reinvent the plumbing. AI routing, LLM observability, and vector retrieval are commodity infrastructure now — most teams rent them from providers who already run the eval loops, gateway logic, and index sharding at scale rather than building from scratch.

Prices verified July 2026. No affiliate links used.

To ensure your directory listings and entity data can be properly extracted by these retrieval databases, test your schema using our Free JSON-LD Entity Graph Visualizer.

The Instability Problem (and Why It Kills the Obvious Strategy)

Your first instinct is to extract every fan-out query and build a page for each. The data says don't.

Surfer analyzed repeated runs and found only about 27% of fan-out queries stay stable across searches, with ChatGPT, AI Mode, and Perplexity each generating different lists for the same prompt. Chasing individual query strings means chasing a target that moves every time you measure it.

The stable layer is the facet, not the string. "HubSpot vs Pipedrive" and "Pipedrive vs HubSpot for small teams" are one facet: the head-to-head comparison. Cover the facet thoroughly and you're eligible for every phrasing the model invents. The same logic explains why a durable content strategy has to be platform agnostic across ChatGPT, Gemini, and AIO rather than tuned to any one engine's current fan-out behavior.

Why Fan-Out Coverage Pays

This isn't theoretical. Surfer's analysis of 173,902 URLs found pages ranking for fan-out queries were 161% more likely to appear in Google's AI Overviews.

Two mechanics drive the payoff. Visibility is fragmented: you can lose the head query and still get cited for owning one facet, which means fan-out is an entry point for sites that could never crack position one.

Authority is cumulative: models favor domains that cover a topic and its facets together, and a site answering five of the twelve sub-queries can get cited multiple times in a single response. Facet coverage compounds; isolated pages don't. This is the same shift that put source mention overlap ahead of domain authority as the retrieval signal that matters.

The Format Advantage: Why Lists, Comparisons, and Directories Win Fan-Outs

Look back at the facet table. Four of the six facet types are structurally list-shaped or comparison-shaped queries.

That's a format signal nobody in the current coverage talks about. When a model runs "top X for Y" as a sub-query, roundups, comparison tables, and directory-style pages are the natural retrieval targets, because their structure matches the question's structure.

The practical moves follow directly. Publish genuine comparison pages for the head-to-heads in your topic, structure roundups with consistent per-item fields (price, best-for, standout limitation) so extraction is trivial, and if you run listing or category pages, treat them as fan-out assets: a well-structured category page is a pre-built answer to every "list of X" sub-query in its niche. This is why human-curated business directories get cited for list-facet queries where individual brand pages don't.

The honesty requirement is real, though. Thin, undifferentiated listicles lose extraction battles to pages with specific, verifiable per-item facts, so the format only wins when the data inside it does.

Writing for Retrieval: the Step-3 Fix

Getting retrieved and getting extracted are different hurdles. Pages fail step 3 when the answer to a sub-query is smeared across the whole document instead of sitting in one liftable chunk.

Four rules cover most of it. Make every section self-contained: a reader (or model) landing on that H2 alone should get a complete answer without context from elsewhere on the page.

Lead each section with the answer, then support it. The definitional sentence goes first; the nuance, evidence, and examples follow, which is the inverse of how most writers build to a conclusion.

Keep one idea per section and give it a heading that mirrors how the sub-query would be phrased. And put concrete facts in the chunk: numbers, names, and dates anchor extraction, and Surfer's data ties fact density to a visibility lift of up to 25% in AI responses.

Topic Clusters Still Do the Heavy Lifting

Facet coverage at scale is just topical authority with a new justification. A hub page on the head topic, cluster pages on each stable facet, and semantic internal links between them make your domain eligible for whichever sub-queries fire on a given day.

The fan-out lens changes cluster planning in one way: you now have ground truth. Instead of guessing subtopics from keyword tools, pull real fan-out queries with Method 1, group them into facets, and build the cluster against observed model behavior.

Prioritize facets where the grounding chunks show weak incumbents. If every retrieval for a facet cites the same three thin pages, that facet is an open net.

When Fan-Out Doesn't Fire

Simple factual queries skip the pipeline. "Capital of Ireland" needs no decomposition, and the model answers from memory or a single lookup.

That boundary matters for planning. Definitional and quick-fact content still competes in the classic snippet game, while comparative, multi-criteria, and advice queries are where fan-out optimization earns its keep. Audit your topics against that line before rebuilding anything.

Fan-Out Meets GTM AI

Query fan-out is the retrieval side of a broader shift on the discovery side, which is the whole point of a GTM AI stack: models reason across many signals in parallel to make one decision. If you sell to buyers using ChatGPT, Gemini, or AI Mode to shortlist vendors, your pages need to survive the same fan-out mechanics your prospects' models use to evaluate you.

Measuring Coverage Without an Enterprise Platform

You can run a workable measurement loop with the free methods alone. Pull fan-out queries for your ten priority prompts monthly via the Gemini script, log which domains appear in the grounding chunks, and track your citation share per facet over time.

Add manual spot-checks in ChatGPT and Perplexity for the same prompts, since each platform fans out differently. When the manual loop starts eating hours, that's the point where paid AI visibility trackers earn their subscription, not before.

Frequently Asked Questions

What is Query Fan-Out in Simple Terms?

Query fan-out is when an AI search system breaks your single question into multiple sub-queries, searches them all at once, and merges the results into one answer. Your content can get cited for a sub-query you never targeted.

How Many Queries Does AI Mode Run Per Search?

It varies by query complexity. Google AI Mode sometimes discloses the count in its interface, and complex questions commonly trigger a dozen or more simultaneous searches. Simple factual queries may trigger few or none.

Can I See the Exact Fan-Out Queries AI Generates?

Yes. The Gemini API returns real fan-out queries in its grounding metadata, the Keyword Surfer extension exposes ChatGPT's searches, and AI Mode discloses search activity in its interface. No paid tool is required to start.

Should I Create a Page for Every Fan-Out Query?

No. Fan-out queries are unstable: research shows only about a quarter stay consistent across repeated runs. Optimize for the stable facets behind them, such as comparisons, pricing, and reviews, through topic clusters and self-contained page sections.

Does Query Fan-Out Replace Keyword Research?

It extends it. Keyword research still identifies the head topics people search. Fan-out analysis adds the layer of sub-questions AI generates on the user's behalf, which reveals facets your keyword tools never surface.

Do ChatGPT, Google AI Mode, and Perplexity Fan Out the Same Way?

No. Each platform generates different sub-queries for the same prompt, and their counts and phrasing differ per run. Cross-platform visibility requires covering the facets they share rather than chasing any single platform's query list.