Short answer: Google AI Overviews SEO is ordinary Google SEO plus a few new controls. To be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to appear in Google Search with a snippet; Google says there are no additional technical requirements, no special schema and no AI text files to add. What actually moves the needle is non-commodity content that answers the sub-questions Google's "query fan-out" searches for, and two Search Console features most sites have not opened yet: the Generative AI performance report and the Search generative AI control.
How do Google AI Overviews and AI Mode choose which pages to cite?
Both features are built on Google's normal Search index and ranking systems, not on a separate crawl. Google's AI features and your website documentation says AI Overviews and AI Mode "may use a 'query fan-out' technique", issuing multiple related searches across subtopics and data sources, and then show a wider and more diverse set of supporting links than a classic results page.
That one mechanism explains most of what you need to do. When someone in Pune types "is term insurance worth it for a 30 year old with a home loan", Google may also search for premium ranges, claim settlement ratios, how a home loan changes the cover amount, and riders. A page does not need to rank for the full question to be cited. It needs to be the best indexed answer to one of those sub-questions.
Third-party data points the same way. Ahrefs' March 2026 study of 863,000 keyword results and 4 million AI Overview URLs found that only 37.9% of cited URLs also ranked in the top 10 for the same query, and 31% did not rank in the top 100 at all. Its earlier July 2025 analysis had put the top-10 share at about 76%. Ahrefs attributes the shift to Google drawing more from fan-out searches. Google does not expose those sub-queries, so treat this as a well-supported explanation, not a confirmed formula.
The practical takeaway: classic rankings still matter (Googlebot has to find and index you), but a focused page that answers a specific sub-question better than anyone else has a real chance of being cited even when a big portal owns position one for the head term.
What do Google AI Overviews SEO guidelines require of a page?
Google's stated requirement is short: the page "must be indexed and eligible to be shown in Google Search with a snippet". Its generative AI optimisation guide adds that the site must also be included in Search generative AI features in Search Console, which is the default setting. Everything else on this checklist exists to make sure those conditions are really true for your important URLs.
| Check | Pass condition | How to verify |
|---|---|---|
| Crawl access | Googlebot allowed in robots.txt and not challenged by the CDN, firewall or hosting bot protection | robots.txt review; Search Console URL Inspection "Live test" |
| Indexed | URL Inspection shows "URL is on Google" | URL Inspection; Pages report |
| Snippet allowed | No nosnippet, no tiny max-snippet, no data-nosnippet wrapped round the main answer | View source and HTTP headers (X-Robots-Tag) |
| Generative AI control | Property set to "included" (the default) | Search Console > Settings > Search generative AI |
| Canonical | Google-selected canonical equals your intended URL | URL Inspection, "Google-selected canonical" |
| Text availability | The key answer, prices and tables are in HTML text, not only inside images or PDFs | Rendered HTML in URL Inspection |
| Discoverability | At least two contextual internal links from related pages | Site crawl or internal links report |
| Structured data | Markup matches the visible text | Rich Results Test |
| Local and product data | Business Profile and Merchant Center details current | Profile and feed review |
The last three rows come from Google's own best-practice list for AI features. The deeper crawl and rendering checks are in our AI search technical SEO audit.
What does Google say you can ignore for AI Overviews SEO?
In May 2026 Google published a dedicated guide to optimising for generative AI features (last updated July 2026). It describes "AEO" and "GEO" as terms people use, but its position is that optimising for generative AI search "is optimizing for the search experience, and thus still SEO". It then lists things you do not need to do. These are worth knowing because agencies still sell several of them.
| Common advice | Google's position | What to do instead |
|---|---|---|
| Add llms.txt or other "AI files" | Not needed; Google Search doesn't use them; neither helps nor harms | Fix indexing first. See does llms.txt help SEO? |
| Add special "AI schema" | Structured data isn't required; no special schema.org markup exists for this | Keep normal, accurate markup. See structured data for AI search |
| Chop content into tiny chunks | No requirement; no ideal page length | Use headings so each section stands alone, at whatever length the topic needs |
| Rewrite everything "for AI" | Not needed; systems understand synonyms and meaning | Write for the buyer; cover their real sub-questions |
| Chase mentions across the web | Inauthentic mentions "isn't as helpful as it might seem"; AI features rely on spam-fighting systems too | Earn genuine coverage with data or expertise others want to cite |
The same guide notes that generative AI features depend on Google's spam-fighting systems as well as its ranking systems, and warns that no third-party tool has access to its internal ranking or AI systems.
How do you optimise content for Google AI Overviews and AI Mode?
Google calls unique, non-commodity content "probably the most important" factor for long-term presence in generative AI search. Commodity content is what anyone could write from common knowledge. Non-commodity content carries your own experience, data or judgement. Here is the process we use for AI Overview and Google AI Mode optimisation; the same work serves both, because Google says both features may use the same query fan-out technique.
1. Map the fan-out for each money query
Take one commercial query and list the sub-questions a careful buyer would need answered. Example for a Noida developer: "3 BHK in Noida Extension under ₹80 lakh".
| Likely sub-question | Page or section that should answer it | Evidence that makes it citable |
|---|---|---|
| What do 3 BHKs actually cost there? | Project page price table | Carpet area, base price, PLC, parking and GST shown as line items with a "prices as of" date |
| Is the project RERA registered? | Project page header | UP RERA registration number in text, linked to the state register |
| When is possession? | Construction update section | Dated construction photos and the RERA-declared completion date |
| How is connectivity? | Location section | Named metro stations and travel times you measured |
| Builder track record? | About/delivered projects page | List of delivered projects with handover years |
Google does not publish its fan-out queries, so this is an informed model of buyer intent, not a list Google has confirmed. It still tells you which facts must exist on your site in plain text.
2. Write answer-first sections
Open each H2 with a two- to four-sentence answer that would make sense if quoted alone, then give conditions and exceptions. An illustrative line such as "Most 3 BHK units in this project cost ₹74–79 lakh all-inclusive (prices as of October 2026)" can be lifted and attributed. "Prices are competitive and value-for-money" cannot.
3. Add evidence only you have
First-party numbers with context are the clearest form of non-commodity content. When we describe healthcare lead generation, we cite the ~₹40 average cost per lead we achieved for Advancells in a restricted ad category, not a generic "healthcare CPLs vary" line. Your equivalent might be average site-visit-to-booking rates, real delivery timelines or a pricing table with dates.
4. Support text with media and keep it current
Google's list of best practices includes supporting text with high-quality images and video. Original photos, short explainer videos and diagrams also help human readers. Update facts when they change and change the visible "updated" date only when the content materially changes.
The wider editorial method, including which formats AI engines cite most, is in how to create content AI search engines cite.
How do you control or opt out of AI Overviews?
Several controls now exist, and they are often confused. This table is built from Google's Search generative AI control help page, the robots meta tag documentation and the AI features page.
| Control | Effect on AI Overviews and AI Mode | Effect on classic results | Effect on AI training |
|---|---|---|---|
| Search generative AI control (Search Console, Settings) | Excluded within 1–2 days; also covers Discover's generative AI features | Google says it is not used as a ranking or inclusion signal elsewhere in Search | None (use Google-Extended) |
nosnippet | Content not used as direct input | No text snippet shown | None |
max-snippet:[n] | Limits how much content is used as direct input | Limits snippet length | None |
data-nosnippet | Marked sections excluded from snippets | Same | None |
noindex | Page not shown | Page removed from Search entirely | None |
| Google-Extended in robots.txt | Not an AI Overviews opt-out | No effect | Limits AI training and grounding in some of Google's other systems |
The control and the generative AI report both rolled out to all websites worldwide as of 31 August 2026. Two practical warnings. First, a site-wide nosnippet added to "protect content" also removes you from AI Overviews, so check SEO plugin and theme settings, which can apply snippet limits site-wide. Second, blocking Google-Extended does nothing for AI Overviews, so it is not a way to opt out of them.
How do you measure Google AI Overviews performance?
Search Console now has a Generative AI performance report for Search. According to Google's help page, it shows impressions only, defined as how many times links to your site were shown in a generative AI feature, covering AI Overviews and AI Mode, with breakdowns by page, country, device and date. It does not show clicks, CTR, position or queries, and it may not appear for sites with too few impressions. Clicks from these features are counted in the normal Performance report under the "Web" search type, alongside everything else.
| Question | Where to look | Limitation |
|---|---|---|
| Which pages appear in AI features? | Generative AI performance report, Pages tab | Impressions only; aggregated per property per feature |
| Is overall organic demand holding? | Performance report, Web search type | AI feature clicks are not separated |
| Do AI-visible pages produce leads? | GA4 landing-page report filtered to organic, plus CRM stage data | Needs clean form and call tracking |
| Are we cited for our key questions? | A fixed monthly prompt panel | Answers vary by user, place and time |
Set expectations on clicks. Pew Research Center's analysis of 68,879 Google searches by 900 US adults in March 2025 found users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% without one, and clicked a link inside the summary in just 1% of visits. Google, for its part, says clicks from results pages with AI Overviews are "higher quality", meaning people spend more time on the site. Both can be true: fewer visits, better visitors. That is why we judge AI Overview work on qualified leads, not sessions. Our AI search visibility tracking framework covers the GA4 and prompt-panel setup.
What is different about AI Mode in India?
India is one of Google's priority markets for AI search. Per Google's India blog (8 October 2025), AI Mode launched in India in English and Hindi, then added Bengali, Kannada, Malayalam, Marathi, Tamil, Telugu and Urdu. India was also the first country outside the US to get Search Live, the voice and camera mode inside AI Mode, and Google describes Indian users as power users of voice and visual search. It adds that AI Mode questions are nearly three times longer than traditional searches globally.
- Write for long, spoken questions. "Which coaching in Kota is good for NEET repeaters with hostel" is a normal AI Mode query. Pages need the specifics such a question asks about: batch types, fees, hostel options, results with year.
- Consider regional-language pages where your buyers use them. A Marathi or Tamil page that is a genuine translation with local details is non-commodity content in that language; machine-translated duplicates are not.
- Keep the Business Profile exact. Local AI answers lean on Business Profile data, which Google lists as a best practice.
A 90-day plan to optimise for AI Overviews and AI Mode
| Period | Work | Done when |
|---|---|---|
| Days 1–15 | Confirm the generative AI control is "included"; audit snippet directives, canonicals and crawl blocks on your 20 most valuable URLs; open the Generative AI performance report and export a baseline | All 20 URLs indexed, snippet-eligible and canonical |
| Days 16–45 | Map fan-out sub-questions for 5–10 money queries; rewrite the matching pages answer-first; add dated prices, specs, processes and first-party numbers | Every sub-question has a section with a quotable answer |
| Days 46–75 | Fill gaps with focused supporting pages; add contextual internal links; refresh Business Profile and Merchant Center data | No priority page has fewer than two internal links |
| Days 76–90 | Compare AI impressions per page against baseline; check leads from those landing pages; run the prompt panel | A report that ties AI visibility to enquiries, not screenshots |
For how this fits with ChatGPT, Perplexity and Copilot, read our AEO and GEO guide. If you want the audit done for you, our AEO and GEO service starts with exactly the indexing and control checks above.
Sources
- Google Search Central: AI features and your website: eligibility, no additional requirements, query fan-out, best practices, preview controls, reporting under "Web", click quality.
- Google Search Central: Optimizing your website for generative AI features: non-commodity content, myths list, AEO/GEO as SEO, inclusion setting, third-party tools.
- Search Console Help: Generative AI performance report (Search): impressions-only metric, features covered, worldwide rollout on 31 August 2026.
- Search Console Help: Search generative AI control: opt-out scope, default, 1–2 day effect, no effect on ranking or training.
- Google: Robots meta tag specifications: nosnippet and max-snippet effect on AI Overviews and AI Mode.
- Ahrefs: 38% of AI Overview citations pull from the top 10 (March 2026): citation overlap with rankings.
- Pew Research Center (22 July 2025): click behaviour with and without AI summaries.
- Google India blog (8 October 2025): AI Mode languages and Search Live in India.