INSIGHTS — 02 · HOW-TO
How to get cited in Google AI Overviews (they follow the organic top 20)
If you sell to buyers in a specific place, Google’s AI Overview is now the first thing many of them read. It sits above the ten links, it names two to eight sources, and it decides for a large share of searchers whether they ever scroll. This article explains where those sources come from, what we see when we test real buyer questions, and the order of work that gets a business into the panel.
The short version, which the rest of the article defends: the AI Overview is downstream of ordinary ranking. Nearly every page we see cited is already near the top of the results for that question or for one of the narrower questions Google splits it into.
Three Google surfaces, one index
Google now answers in three places, and it helps to keep them apart:
| Surface | Where it appears | How it picks sources |
|---|---|---|
| AI Overview | Above the results for many informational and commercial queries | Search ranking for the query and its generated sub-queries |
| AI Mode | A separate tab, full conversational answers | Broader “query fan-out”: many sub-queries, then synthesis |
| Gemini app | Standalone assistant | Google Search grounding, same index, more conversational |
All three read the same crawl. There is no separate submission, no AI sitemap, no meta tag that opts you in. If Googlebot can fetch and index a page and that page earns a place in the results for the right questions, it is eligible. If not, nothing else you do matters.
One correction worth making early: Google-Extended in robots.txt does not control AI Overviews. Google documents it as a control for Gemini model training and for grounding in the Gemini app. AI Overviews and AI Mode use the ordinary Googlebot crawl. Blocking Google-Extended will not keep you out of AI Overviews, and allowing it will not get you in.
Query fan-out: the question behind the question
Ask Google “who builds commercial greenhouses in southern Ontario” and the AI system does not run that one search. It fans the question out into several: greenhouse builders Ontario, commercial greenhouse cost, greenhouse construction companies near a city, permits for agricultural buildings, and so on. It retrieves top results for each, reads them, and writes one answer.
This is why a page can be cited for a question it never targeted, and why a page that ranks eighth for the exact phrase can be skipped entirely: it did not rank for any of the sub-questions the system actually ran. The pages that get cited are the ones that sit near the top for the cluster, not the single phrase.
For a business this changes the content plan. Instead of one page chasing one keyword, you need a small set of pages that each answer one sub-question well, linked together, so that whichever way Google fans the buyer’s question out, one of your pages is in the top results for it.
What we see when we test buyer questions
We put real buyer questions to Google’s AI surfaces on behalf of clients and record every source cited. The specifics stay with the clients; two general patterns are worth stating.
The more specific and place-bound the question, the more likely a local company is named. Broad questions with no place attached are answered from publishers, associations and national directories. Questions qualified by a region, a city or a concrete situation are where individual businesses appear, usually with their own site cited. The realistic route in is to own the qualified questions, not the generic ones.
Third-party pages carry weight. Chamber pages, regional news and association listings rank for the sub-questions on their own, and when they name you, the engine has two sources saying the same thing. A business with no third-party footprint has to do all of the ranking work alone.
Citations vary between runs. The same question on the same day can cite a page once and skip it the next time. Any single check is anecdote. We treat a source as established only when it repeats across runs and weeks.
The order of work
1. Make sure Googlebot can read every important fact
Fetch your key pages with JavaScript disabled. If the service list, the service area, the pricing or the contact details disappear, the engine never saw them. Move them into HTML. Check robots.txt and your CDN’s bot settings for anything that blocks Googlebot on those paths. Confirm the pages are indexed in Search Console, not just submitted.
2. Own the qualified questions in ordinary search
Write down the questions your buyers ask, with the place or situation in them. For each, check where you rank today. The realistic targets are the ones where you already appear on page one or two; those are the sub-questions where a better page can move you into the citation pool. Build or rewrite one page per question: the question as the heading, the answer in the first screen, then the detail.
3. Write answer-first, with facts an engine can lift
The cited passage is usually a sentence or two that directly answers the sub-question with a concrete fact: a price range, a lead time, a service radius, a certification. Put that sentence near the top of the page, in plain text. Tables help: a table of models and prices, or of areas and delivery times, is the easiest thing for the system to quote.
4. Keep the facts consistent everywhere
The engine cross-checks you against your Google Business Profile, your structured data and the directories it finds. If the service area on the page says one thing and the profile says another, the citation goes to someone whose facts agree. Make name, address, phone, category and area identical across all of them, and add LocalBusiness or Organization, Service and FAQPage structured data that repeats the same facts.
5. Get corroborated
Association member pages, chamber listings, supplier partner pages, a local news mention. These rank for the sub-questions on their own, and when they name you, the engine has two sources saying the same thing. One good third-party page can do more than three new pages of your own.
6. Measure the way it actually shows up
There is no API for AI Overviews and Search Console does not break them out. The working method: a signed-out browser, the buyer questions, a sheet with three columns (named, cited, who else), repeated monthly. Read trends, not single answers. Gemini can be sampled the same way and tends to move earlier than AI Overviews because it is less tied to the ranking of the exact phrase.
What does not work
- Chasing the generic head term. “Modular homes” or “roofing contractor” without a place cites publishers, not vendors. The panel for that question is not available to you.
- Adding AI-flavoured text. A paragraph that begins “As an AI-friendly resource” does nothing. The system reads facts, not signals of intent.
- Blocking crawlers to protect content. A site that returns 403 to Googlebot on its service pages has opted out of every Google surface at once.
- Waiting for a submission form. There is none. Ranking for the sub-questions is the submission.
Where to start
If you want to know which of your buyer questions currently produce an AI Overview, whether you are in it, and who is cited instead, the fastest path is a measured baseline. We run twenty real buyer questions across Gemini and ChatGPT, record every source, and send back a PDF within two business days. It is free.
Questions people ask
How does Google choose sources for AI Overviews?
Google has said AI Overviews are built from its core search ranking and use the same crawl as Search. In practice the cited pages are almost always ones already ranking on the first two pages for the question or for the sub-questions Google generates from it. There is no separate index to submit to.
Does blocking Google-Extended remove me from AI Overviews?
No. Google documents Google-Extended as a control for Gemini model training and Gemini app grounding. AI Overviews and AI Mode use the ordinary Googlebot crawl. Blocking Googlebot removes you from everything; blocking Google-Extended does not affect AI Overviews.
Can I see AI Overview citations in Search Console?
Only indirectly. Google counts AI Overview impressions and clicks inside the normal Search performance report without a separate filter. The practical check is to search the question yourself in a signed-out browser and read the sources panel, then repeat on a schedule.
Why does the AI Overview cite a big national site instead of a local company?
Because the question was generic. For broad questions Google cites publishers, associations and national directories. Local and regional companies are named when the question is qualified by place or situation. Owning the qualified questions is the realistic route in.
( WHERE YOU STAND )
Does AI recommend you?
We put twenty real buyer questions about your category and market to Gemini and ChatGPT and send you a PDF showing who they cite instead of you, plus fourteen technical checks. Free, no call required.
Request the free audit