How to Rank in Google AI Overviews: A Step-by-Step Playbook (2026)

By the AEOeye editorial team·Updated Oct 3, 2026

Part of our pillar guide: Getting Recommended by AI Engines

The short answer

To appear in a Google AI Overview or AI Mode, a page must be indexed and eligible to show a snippet. Google names no other requirement and says no special markup, files or AI-only rewrites are needed. Getting cited then comes down to ranking work: cover the sub-questions behind the query, answer each first-hand under its own heading, and track impressions in Search Console.

You don't rank in an AI Overview the way you rank in blue links. Google builds each summary while the search runs and picks supporting links from its own index, so there is nothing to submit, buy or lock in. Google's own guide calls optimizing for it "still SEO". We agree, with one twist: you rank for the sub-question, not the keyword.

We haven't run controlled AI Overview experiments; this guide rests on Google's documentation, published studies and our own crawler data.

What is a Google AI Overview, and how is AI Mode different?

An AI Overview is an AI-generated summary shown with some results, linking to supporting pages. AI Mode is a separate Search experience for questions that need more exploration or comparison. Both draw on Google's index, share the same eligibility rules and may use query fan-out: several related searches across subtopics (Google Search Central). Which queries trigger an Overview is covered in what triggers one.

A close-up of a hand holding a smartphone with Google search displayed on the screen

What must be true before any tactic can work?

The page must be indexed and eligible to appear with a snippet. Google lists no additional requirement, and even a page that meets every requirement may not be indexed or served (Google's guide, updated July 2026). Check three things:

Crawlable: robots.txt doesn't block Googlebot, and key facts sit in rendered text, not only in images or behind a login.
Indexed: Search Console's URL Inspection tool reports it as indexed.
Snippet-eligible: no noindex, and no nosnippet, tiny max-snippet or data-nosnippet on the key passage. nosnippet also blocks direct use in AI Overviews and AI Mode (Google).

Google-Extended is a different control. Google says it doesn't affect inclusion in Google Search (Google) and points to the controls above, plus Search Console's site-wide Search generative AI control, for limiting AI features (Google). In AEOeye's crawler access study (October 2026, a convenience sample of 384 well-known sites), 31 sites (8.1%) block it, including 20 of 38 news publishers, yet by Google's documentation that doesn't decide whether they appear in AI Overviews.

Is ranking in organic results still the price of admission?

Partly. Google says its AI features are rooted in its core ranking and quality systems, so indexing and relevance still decide who is considered. But the top spot for the head keyword isn't the ticket: Google says its models find extra supporting pages, so AI features link to a wider, more diverse set of pages than classic results (Google).

Two vendor studies show how loose the link is:

BrightEdge (September 2025, nine industries): 54.5% of AI Overview citations also ranked organically, up from 32.3% in May 2024, yet only 16.7% came from the top 10. Overlap was 75.3% in healthcare but 22.9% in e-commerce, so rankings matter most where trust does (BrightEdge).
Ahrefs (March 2026, 863,000 keyword SERPs): 37.9% of cited URLs ranked top 10, 31.2% ranked 11 to 100 and 31.0% beyond 100, against about 76% in the top 10 in July 2025 (Ahrefs).

Neither is Google data; BrightEdge states no sample size or rank cutoff, and Search Engine Journal reports Ahrefs improved its parsing between studies (SEJ), so don't compare them. Ahrefs explains the shift with fan-out: Overviews lean on pages that surface in sub-question results. Our reading: a page can rank first for the head query and be skipped, while one that ranks well for a sub-question gets the link.

How do you write passages an AI Overview will lift?

Lead each section with a direct answer under a question-style heading, so the passage works on its own. Google says you needn't write or chunk content specially for AI, and it already ranks individual passages in Search (ranking systems guide), so treat this as courtesy to readers, not a hack.

Heading = the question a person would type.
Lead with a 40-60 word answer that includes the number or condition, then one idea per paragraph.
Name the entity: "Search Console's Generative AI performance report shows impressions", not "it shows impressions".
Add what only you have: your data, screenshots, method. Google says original, helpful content will likely matter more in the long run than any other tip in its guide.

Distrust "ideal passage length" numbers and "AI Overview scores": Google's guide gives none. The peer-reviewed GEO study (KDD 2024) found that statistics, quotations and citations raised visibility, but it didn't test AI Overviews (arXiv).

Do schema, llms.txt or E-E-A-T help?

Not as levers. Google says no special schema is needed, its Search ignores llms.txt and similar files, and E-E-A-T (experience, expertise, authoritativeness, trust) isn't itself a ranking factor (Google's guide; Google).

Schema: keep it accurate and generated from the same data as the visible page, as ours is for FAQs. It still supports rich results, but 74.7% of the 312 well-known-brand homepages we parsed already carry JSON-LD, so it rarely differentiates. See structured data for AI.
llms.txt: Google says it's fine to publish one for other systems. We do, but it won't get you into an AI Overview.
E-E-A-T: trust matters most. Show who wrote the page and cite primary sources. Never invent authors or credentials; Google calls fabricated creator profiles deception. Earn real mentions; chasing inauthentic ones is less helpful than it looks.

How do you cover the whole question, not just the keyword?

Treat each query as a cluster of sub-questions and answer each one on a single page. Google's example: "how to fix a lawn that's full of weeds" fans out into searches like "best herbicides for lawns" and "remove weeds without chemicals" (Google).

List the sub-questions: People Also Ask, related searches, the topics the Overview covers and the queries Search Console already shows. Google doesn't publish its real fan-out queries, so these are proxies.
Check who ranks for each. Wikipedia, YouTube and Reddit were 15% of the sources in Pew's March 2025 AI summaries (Pew); where they dominate, find another angle.
Give each sub-question its own heading and answer. Same-intent questions share a page; a different intent gets its own, linked. Google says separate content for every phrasing, made mainly to manipulate AI responses, violates its scaled content abuse policy.
Keep time-sensitive facts current, and change a visible date only when the content really changed (Google).

We follow the one-page rule ourselves: on 2026-10-03 we merged 165 duplicate-intent pages into 37, and this guide replaces two earlier pages of ours. We haven't measured the effect on AI Overview citations yet, but we bet one strong page per question beats five thin ones.

Close-up of a hand writing in a notebook with a checklist

How do you track whether you're cited?

Use Search Console's Generative AI performance report for impressions, then spot-check queries by hand and repeat weekly. The report counts how often links to your site appeared in AI Overviews and AI Mode.

Google's help page describes impressions only and says the report reached all websites worldwide as of August 31, 2026, though sites need enough impressions to see it (Search Console Help). AI-feature clicks stay inside the standard Performance report's Web totals (Google).

Spot-check 20 to 30 fixed queries while logged out, from one location, and record whether an Overview appears, whether you're cited and who is cited instead. Repeat weekly: Ahrefs (November 2025) tracked over 43,000 keywords and found a 70% chance an Overview changed between observations, with only 54.5% of cited URLs shared between consecutive responses (Ahrefs).

Judge by impressions and branded search, not clicks: in Pew's March 2025 data, people clicked a link inside an AI summary in just 1% of visits (Pew). See AI Overviews and your traffic.

AEOeye answers a different question: do AI engines name your brand when buyers ask? The free audit tests ChatGPT; the $29 one-time report adds Google AI, Gemini, Claude and Perplexity. For scheduled tracking of many queries, a dedicated tracker fits better: Profound and Otterly.AI both name Google AI Overviews on their sites as of October 2026; we haven't tested either.

A professional workspace with computers and analytical graphs on a monitor

Key terms

AI Overview: A Google-generated summary shown with some search results, with links to supporting pages. To appear as a link, a page must be indexed and eligible to show a snippet. ↗
AI Mode: A separate Google Search experience for questions that need more exploration, reasoning or comparison. It follows the same eligibility rules as AI Overviews. ↗
Query fan-out: A technique where Google's models issue several related searches across subtopics and data sources to build a response, such as turning a lawn-weed question into herbicide, chemical-free and prevention searches. ↗
E-E-A-T: Experience, Expertise, Authoritativeness and Trust: Google's framework for judging content quality, with trust the most important element. Google says it is not itself a ranking factor. ↗

Step-by-step

1
Confirm the page is eligible
In Search Console's URL Inspection, check that the page is indexed, that robots.txt allows Googlebot and that no noindex, nosnippet, low max-snippet or data-nosnippet limits it. Google lists indexing plus snippet eligibility as the requirement.
2
List the sub-questions behind the query
Write down the searches a person would run next: People Also Ask, related searches, the topics the Overview covers and the queries Search Console already shows. Google doesn't publish its real fan-out queries, so these are proxies.
3
Check who ranks for each sub-question
Read the top results for each one. Keep the sub-questions where you can add original data or first-hand experience, and skip those dominated by Wikipedia, YouTube or Reddit unless you have a new angle.
4
Answer each one first, under its own heading
Use the question as the heading, then give a direct 40-60 word answer that names its subject and includes the number or condition, followed by evidence. Keep same-intent sub-questions on one page rather than one page per phrasing.
5
Add what only you can add
Include your own data, screenshots, method or hands-on experience, and link a primary source for every statistic. Google says original, helpful content will likely matter more over time than any other tip in its guide.
6
Keep markup honest and skip the myths
Keep structured data accurate and identical to the visible text. Don't spend time on llms.txt, content chunking or AI-only rewrites for Google, which says its Search doesn't need them.
7
Keep time-sensitive facts current
Update prices, dates and statistics when they change, and change the visible date only when the content really changed. Google runs freshness systems that surface newer content for queries that need it.
8
Measure impressions, then repeat
Check Search Console's Generative AI performance report for impressions, spot-check a fixed query list while logged out, and repeat weekly because AI Overviews change often. Track your citation rate: queries where you're cited divided by queries where an Overview appears.

	What Google or the data says	Our call
llms.txt and other special AI files	Google says its Search doesn't use them; they neither help nor harm	Skip for Google; harmless to keep for other tools
Chunking pages into tiny pieces	Google says there is no requirement and its systems surface the relevant piece	Skip
Special schema for AI Overviews	Google says structured data isn't required and no special markup exists	Keep it accurate for rich results; expect no lift
Blocking Google-Extended to opt out	Google says it doesn't affect inclusion in Google Search	Doesn't opt you out; use Search Console's Search generative AI control or snippet controls
Ranking first for the head keyword	37.9% of cited URLs ranked top 10 in Ahrefs' March 2026 data	Rank for the sub-questions too
First-hand, original content	Google says it will likely matter more over time than its other tips	Do this first
Counting clicks as success	Pew: 1% of visits clicked a link inside an AI summary	Count impressions and branded search

Key takeaways

Eligibility is the only hard gate: indexed and snippet-eligible. Skip llms.txt, chunking and "AI Overview scores" for Google.
Cited pages often rank outside the top 10 for the main query, so rank for the sub-questions.
Give each sub-question an answer-first section with first-hand proof, on one strong page.
Track impressions in Search Console and repeat spot-checks; Overviews change often and cited links rarely get clicks.

See how AI talks about your brand

Run a free AI visibility audit in under a minute.

FAQ

Do I need to rank in the top 10 to be cited in an AI Overview?+

No, but ranking for the related searches helps. Ahrefs found 37.9% of cited URLs ranked in the top 10 for the original query and 31.0% ranked beyond position 100 (March 2026, 863,000 keyword SERPs). Ahrefs attributes this to Overviews pulling from sub-question results, so aim to rank for the sub-questions, not only the head term.

Do I need schema markup, llms.txt or special formatting?+

No. Google says structured data isn't required, no special schema exists, and its Search doesn't use llms.txt or similar files. Chunking and AI-only rewrites aren't required either. Keep structured data accurate and matching the visible text for rich results, and write answer-first because it helps readers.

Does blocking Google-Extended keep me out of AI Overviews?+

No. Google says Google-Extended doesn't affect inclusion in Google Search and isn't a ranking signal; it limits training of Google's models. To keep a whole site out of AI Overviews and AI Mode, use Search Console's Search generative AI control, rolled out to all sites on August 31, 2026; Google says it doesn't affect other parts of Search. Page-level nosnippet, max-snippet and data-nosnippet also work but shrink your regular snippets. In AEOeye's October 2026 study, 31 of 384 sites blocked Google-Extended.

Can I pay or submit a page to get cited in AI Overviews?+

Google's documentation describes no such route. It lists indexing and snippet eligibility as the requirements, says no other special optimizations are needed, and warns that indexing and serving aren't guaranteed even when a page meets every requirement. Treat any offer of guaranteed AI Overview placement with suspicion.

How do I know whether my pages appear in AI Overviews?+

Open Search Console's Generative AI performance report for impressions in AI Overviews and AI Mode by page, country, device and date. Then spot-check a fixed list of queries while logged out and repeat weekly. Google says the report reached all websites worldwide as of August 31, 2026, but sites need enough impressions to see it.

Sources

Guides

How to Add an llms.txt File: Create, Host and Validate It (2026)

Read Guides

How to Fix AI Getting Your Brand Wrong (Old Prices, Wrong Features)