
Someone asks Google AI Mode which payroll software suits a 40-person company with staff in two countries. Google doesn’t run that sentence as one search. It splits the question into subtopics and runs a batch of searches at the same time. Query fan-out is that split, and the pages that answer the sub-queries are the ones the final answer can draw on.
Step 3 is where most of the surprise comes from. Google’s Robby Stein gave an example in a July 2025 interview reported by Search Engine Journal. For a group trip, AI Mode may think up questions about restaurants and bars, then search those too.
Google named the technique when it introduced AI Mode in March 2025. The announcement says AI Mode issues “multiple related searches concurrently across subtopics” and then brings the results together into one response.
At Google I/O in May 2025, the company added Deep Search to AI Mode. It applies the same method at a larger scale and can issue hundreds of searches for one question.
Google’s AI features documentation for site owners confirms three things:
The same page notes that AI Mode and AI Overviews may use different models and techniques. So the links each one shows for the same question will vary.
These three get mixed up, and many engines use all of them for one prompt.
| Technique | What happens | Timing | Example |
|---|---|---|---|
| Query expansion | One query is rewritten or enriched with synonyms and context | Before the search | “payroll tool” becomes “payroll software for small businesses 2026” |
| Query fan-out | One question becomes several searches on different subtopics | In parallel | Separate searches for compliance, pricing and reviews |
| Multi-step retrieval | Each new search is chosen after reading the last results | In sequence | Find vendors first, then open each vendor’s pricing page |
| Research modes | Long runs that mix fan-out and follow-up rounds | Often minutes | Deep Search, which can issue hundreds of searches |
Expansion changes the wording of a single search. Fan-out multiplies the searches. Multi-step retrieval adds a feedback loop, where what the system reads shapes what it looks for next.
Google is the only one that documents it by name. Google AI Mode was built around it. AI Overviews may use it, and Deep Search runs it at the largest scale.
Other engines behave the same way without the label. OpenAI says ChatGPT search rewrites a prompt into one or more targeted queries before it calls search partners. ChatGPT, Perplexity and Microsoft Copilot often send several searches for one prompt, and those sub-queries show up in tracked answers.
Microsoft has started to expose part of this. The AI Performance report in Bing Webmaster Tools lists grounding queries, the phrases Copilot used to retrieve content.
Here’s how the payroll question from the opening could fan out. The sub-queries are illustrative, but they follow the pattern engines show in tracked answers.
The prompt never said “employer of record” or “2026”. A vendor with a sharp cross-border payroll page can be pulled in through the first sub-query. It doesn’t need a page about “40-person companies”. A vendor missing from the roundups that rank for the fourth sub-query may never reach the shortlist.
It opens more doors. Each sub-query is a separate chance to be retrieved. That helps explain why long-tail queries and facet pages can earn citations for prompts they never target.
It also raises the bar for being cited. Fan-out widens what an engine reads, but the list it shows stays short. In our tracking, Google AI Mode answers carried about 4.8 links each, against about 13.7 for AI Overviews. Source: Ansvisor tracking data, September 15 to October 14, 2026.
Many pages get read and few get shown. The ones that win answer one facet in a passage the model can lift whole.
And it changes who your competitors are. A brand that never ranks for your main prompt can still win the pricing or integration sub-query and take the citation there. This is why prompt coverage measured on the top-level prompt alone can hide gaps.
A pillar page with supporting pages, the structure described under topic clusters, maps well onto fan-out. Our guide on finding AI search opportunities shows how to turn sub-queries into a content plan.
Google doesn’t publish them. The generative AI report in Search Console shows impressions in AI Overviews and AI Mode by page, but no queries. Bing’s grounding queries are the closest first-party view, and they cover Copilot only.
The rest comes from tools, and there are two kinds. Some predict likely sub-queries with a language model. Others record the searches engines actually ran while answering tracked prompts. Predictions are useful for brainstorming, while observed queries show what an engine searched.
Ansvisor’s Query Fan-Out view shows the sub-queries engines actually ran for your tracked prompts. They come from ChatGPT, Perplexity, Microsoft Copilot and other engines, observed rather than predicted. The High Frequency view surfaces sub-queries that recur across thousands of prompts, and By Prompt shows how one question was researched. Each sub-query carries an intent label, and one click adds it to Prompt Monitoring so you can track mentions and citations for it. Start a 14-day free trial to see the fan-out behind your own prompts.
Each prompt you win brings more mentions, more citations and more buyers. Ansvisor keeps finding the next one.
Query fan-out is how AI search engines break one question into several sub-queries on different subtopics and run them at once. The results are merged into one answer with supporting links. Google introduced the term with AI Mode in March 2025.
Google says AI Mode issues multiple related searches across subtopics and data sources, then brings the results together. Its Deep Search mode applies the same method at a larger scale and can issue hundreds of searches. AI Overviews may use fan-out too.
Query expansion rewrites or enriches a single query, for example by adding synonyms or context. Query fan-out turns one question into several separate searches on different subtopics. Engines often do both before they write an answer.
Find the sub-queries behind the prompts that matter to revenue and group them by facet, such as pricing, comparisons and integrations. Answer each facet in its own section, near the top. Keep pages indexed and earn mentions on the third-party pages that rank for comparison sub-queries.
Google doesn’t publish them, and Search Console shows no AI Mode queries. Bing Webmaster Tools lists grounding queries for Copilot. Platforms such as Ansvisor record the sub-queries ChatGPT, Perplexity and Microsoft Copilot actually ran for tracked prompts.
See how AI answers describe your brand, find out why competitors get picked, and close the gaps with AI agents that work alongside your team.
14-day free trial or 50% off your first month with code SAVE50.
Continue exploring key AI visibility concepts.
Measure and improve how often your brand appears in AI-generated answers.
Learn more →Strategies for increasing visibility in answer engines and AI summaries.
Learn more →Optimizing content for AI-powered discovery experiences.
Learn more →Understand how OpenAI retrieves and synthesizes information.
Learn more →AI-generated summaries that appear directly in Google Search.
Learn more →Explore how Perplexity cites and presents sources.
Learn more →References and sources used by AI systems to support answers.
Learn more →Measure the quality and influence of cited sources.
Learn more →How easily AI systems can discover and reuse your content.
Learn more →New terms are added regularly.
Help us improve the page or suggest a new term →
Co-founder at Ansvisor
Cihan Geyik is the co-founder of Ansvisor, an open-source AI Visibility platform for AI Search. With more than 15 years of experience in digital marketing and growth, he writes about AI visibility, AI search, AEO, GEO, citations, and answer engines. He focuses on helping brands understand and improve their presence across ChatGPT, Gemini, Google AI Overviews, Bing AI Mode and other AI answer engines.
Add Ansvisor as a preferred source and Google will show you more of our AI search, AEO and GEO guides.
© 2026 Ansvisor. All rights reserved. Ansvisor is an open-source AI Search Execution Platform for AI Visibility. Ansvisor is a brand owned and operated by Empler AI Inc.
The open-source AI Search Execution Platform. Track how AI answers talk about your brand and act on it.
Open source · MIT license© 2026 Ansvisor, a brand of Empler AI Inc. · 1401 PE Ave. Unit 105, Wilmington, DE 19806, United States
Cookie preferences
We use cookies to keep the site working and, with your OK, to measure traffic and improve our marketing. Read our Cookies & Tracking policy.