The essentials
- The mechanism: AI Mode breaks the question into parallel sub-queries (query fan-out) and assembles the answer from the pages selected for each one.
- The consequence: a page gets cited if it answers a sub-question precisely, not if it targets the main query; covering a subject beats optimising a page.
- The format: self-contained sections of 80 to 150 words, opening with the answer, with a verifiable, dated fact.
- Measurement: share of citations across a fixed set of questions, taken quarterly; Search Console doesn't yet separate out these citations.
Google's AI Mode replaces the list of links with a written, conversational answer, citing five to fifteen sources. Its definition and how it works are set out in AI Mode: definition ; this article gives the working method for appearing in it, which differs on several points from classic SEO and from optimising for the AI Overviews.
Understanding what fan-out changes
Asked "which marketing tracking solution for a small manufacturer", the engine doesn't search for that phrase. It breaks it into sub-questions: what a marketing tracking tool is, what the main solutions cost, how they differ, what users say about them, how to set one up. Each queries the index separately, and the final answer assembles passages from different pages. A company can therefore be cited on a question it doesn't target, because it answers one of the sub-questions well. Conversely, a page perfectly optimised for the main query but treating no sub-question in depth won't be selected.
Step 1: map the sub-questions
- List a buyer's questions: definition, price, selection criteria, comparison, implementation, mistakes, alternatives, lead times, expected return.
- Check with customers: the questions actually asked before purchase, noted on the phone or by the customer service team.
- Fill the gaps with tools: Google's "People also ask", the suggestions, and the Search Console queries the site already appears on.
- Test the fan-out: put the main question to an assistant and observe the angles its answer covers; they reveal the sub-questions the engine picked.
- Assign every sub-question: to an existing page or to a section to be written. A sub-question your site doesn't answer is a citation handed to a competitor.
Step 2: write extractable passages
| Element | Rule | Why |
|---|---|---|
| Section heading | The sub-question, phrased the way the user asks it | The engine matches the question with the heading |
| First sentence | The direct answer, with no preamble | This is the passage that gets extracted |
| Passage length | 80 to 150 words, understandable out of context | These systems work on passages, not on pages |
| Factual content | A figure, a range, a date, a source | Verifiable facts are preferred to assertions |
| Tables | For prices, comparisons, specifications | Reused as they stand in comparative answers |
| Vocabulary | The customer's words, not internal jargon | Matching happens on the meaning of the phrasing |
The difference from classic SEO lies in the passage's autonomy: a paragraph beginning "as we saw above" can't be cited. The writing principles are developed in structured content for LLMs.
Step 3: make the expertise verifiable
The engine favours sources consistent with what it already knows. Four things contribute: an identified author with their role and experience, an About page describing the company with verifiable facts, structured data linking the organisation, the people and the articles, and an identical description across every external source. A company whose information contradicts itself from one directory to the next is set aside in favour of a stable source. That work is detailed in optimising your About us page.
Step 4: take care of local and transactional
AI Mode incorporates Google Business Profile listings, product data and reviews into its answers. For a local business, a complete listing, precisely categorised, with hours, photos and recent reviews, feeds directly into the answers to proximity questions. For an online shop, the product feed and the structured price and availability data are read as they stand. These elements count as much as editorial content on the corresponding queries.
Step 5: measure
Search Console doesn't yet separate out citations obtained in AI Mode. Measurement rests on a simple protocol: fix a list of ten to twenty shortlisting questions for your market, put each of them three times at regular intervals, and note whether you are cited and with which page. The ratio of questions where you appear to questions tested is your share of voice. A quarterly reading is enough; the tools on the market are compared in the GEO tracking tools comparison. The complementary metrics are in GEO KPIs.
Mistakes to avoid
- Treating AI Mode as a separate channel: it uses the Google index; without proper indexing and organic ranking, nothing happens.
- Writing for the main query alone: fan-out queries the sub-questions, and that is where the citation is decided.
- Diluting the answer: an answer buried in eight hundred words of introduction doesn't get extracted.
- Multiplying short pages: a pillar page covering ten sub-questions is picked up better than ten three-hundred-word pages.
- Concluding from a single test: the same question produces different answers; only repetition measures anything.
How GreenRed helps
Rather than juggling several tools, GreenRed's GEO tracking brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
How do you get cited in Google's AI Mode?
By covering a subject's sub-questions in depth rather than the main query alone, with sections whose heading repeats the question and whose first sentence gives the answer, in 80 to 150 self-contained words containing a verifiable fact, on a site that is indexed and already ranking well.
What is query fan-out?
The breaking of the user's question into several sub-queries run in parallel against the Google index. The final answer assembles passages from different pages, which is what lets you be cited on a question the page doesn't explicitly target.
Do you need different content for AI Mode and for SEO?
No, but the format shifts: each section must stand on its own and open with the answer. The foundations stay the same — indexing, organic ranking and quality; AI Mode uses the same index as classic search.
How do you measure your visibility in AI Mode?
By fixing ten to twenty of the market's shortlisting questions, putting each of them three times at regular intervals and noting the citations obtained. The ratio of questions where the brand appears to questions tested gives the share of voice, to be taken quarterly.