Answer Engine Optimization for AI Overviews

1. Start with what Google actually says

Most AEO advice in circulation contradicts Google’s own documentation. The relevant page is AI features and your website (Google Search Central, updated 10 December 2025), and it is unusually blunt.

• On requirements: there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary,” and “no additional technical requirements.”
• On eligibility, the only stated bar: a page “must be indexed and eligible to be shown in Google Search with a snippet.”
• On the files and markup this industry sells: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.”
• On how citations are chosen, Google describes query fan-out — issuing multiple related searches across subtopics and data sources — which it says lets Search display a wider and more diverse set of links than a standard results page. This is Google’s own explanation for why AI Overview citations do not mirror the top ten, and it is the single most useful sentence in the documentation.

Google’s best-practice list is ordinary SEO: allow crawling, create internal links for discoverability, provide a good page experience, keep important content in textual form, support text with quality images and video, ensure structured data matches visible text, and keep Merchant Center and Business Profile current.

The practical reading: there is no AI Overview switch to flip. The eligibility gate is snippet eligibility, and everything past that gate is a competition to be the best available source for a sub-query you may never have targeted directly.

2. What triggers an AI Overview

This is the strongest, most reproducible finding in the entire field — two independent methodologies, a year apart, on different populations, reaching the same answer.

A 6.8× difference between question and non-question queries, replicated. By topic, the Washington University study found Hobbies & Leisure 46.1%, Science 39.9%, Health 26.6%, Business & Finance 26.2%, Law & Government 9.6%, Politics 7.5%.

Query form is the strongest observed predictor of AI Overview presence. Interrogative, longer, natural-language queries trigger Overviews; short head terms mostly do not.

How often overall? The studies disagree, and the reason matters

Semrush’s data also shows the intent mix shifting: informational queries fell from 91.3% of AI Overview triggers in January 2025 to 57.1% in October 2025, while commercial rose from 8.2% to 18.6% and transactional from 2.0% to 13.9%. Overviews are moving into the commercial queries that matter most.

3. Who actually gets cited

Here is where practitioner advice most often goes wrong, because two different statistics get quoted interchangeably.

Question A: what share of citations come from a top-10 page?

Question B: what share of Overviews cite at least one top-ranking page?

seoClarity found 90% of AI Overviews showed at least one overlap with the top 10, and 94% with the top 20 — but only 6% showed complete overlap, with an average of three top-20 URLs per Overview, down from five in May 2025.

“90% of AI Overviews include a top-10 page” and “17–38% of citations come from the top 10” are both true. They answer different questions.

The trend, and an honest caveat

Ahrefs’ own two studies show top-10 sourcing falling from 76.1% to 37.1% between July 2025 and January 2026, which Ahrefs attributes to query fan-out. BrightEdge reports its figure as flat at roughly 17% over a similar window. These cannot all describe the same reality.

Part of the gap is methodological: the two Ahrefs studies used different definitions — the top three most-visible citations versus all 4M URLs, and “blocks” versus organic positions. The 76% figure is superseded and should not be quoted as current, including where it still appears in Ahrefs’ own guidance pages.

What survives every study: a substantial share of AI Overview citations come from pages that do not rank in the top 10, and a meaningful share rank nowhere on page one at all — 14.4% not ranking in Ahrefs’ 2025 study, 36.7% outside the top 100 in its 2026 update, 29.8% of cited domains absent from first-page results in the Washington University academic audit.

What the #1 position is worth

• seoClarity: position-1 URLs appear in AI Overviews 43% of the time.
• Originality.AI: the #1 result has a 57.9% citation probability, but accounts for only 8.0% of all citations. A page-one document has roughly 38% probability.
• Ahrefs (July 2025): the median organic rank of the first-listed citation was position 2.

Ranking first is neither necessary nor sufficient. Google omits the #1 result from the Overview roughly half the time.

One more citation pattern worth knowing

Ahrefs (March 2026) found YouTube accounts for 5.6% of all AI Overview citations and 18.2% of non-ranking citations — up 34% over six months. Video is a citation surface, not just a content format.

4. What it costs when an Overview appears

Every credible study agrees on direction. None agrees on magnitude, and the disagreement is instructive.

Ahrefs isolates position 1 on informational keywords; Amsive averages across all positions and includes branded terms. That reconciles most of the −15% versus −58% gap.

The most actionable finding in this group is Amsive’s split: branded queries gained 18.7% while non-branded lost 20.0%, and only 4.79% of branded keywords triggered an Overview at all. Brand demand is substantially insulated from this. That is a strategic argument for building brand search, not just topic rankings.

And note the outlier honestly: Semrush found zero-click rates falling on the same keywords after they gained an Overview. It contradicts the prevailing narrative. We report it rather than omit it.

5. What does not work

Schema markup, as an AI Overview lever

Ahrefs ran a difference-in-differences study on 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 controls. The effect on AI Overview citations was −4.6%, p ≈ 0.0004 — significant and negative. Effects on AI Mode (+2.4%) and ChatGPT (+2.2%) were indistinguishable from zero. Google’s documentation independently states no special structured data is needed. Implement schema for rich results and entity clarity in classic Search; do not sell it as an AI Overview tactic.

FAQ schema, specifically

FAQ rich results were curtailed in August 2023 and fully deprecated in 2026 — removed from results in May 2026, with Search Console API support ending that August. (Verified via Search Engine Journal’s reporting of Google’s statement; we could not retrieve Google’s own announcement page directly.) Marking up FAQs for AI Overview benefit is optimising for a feature that no longer exists.

llms.txt

Ahrefs analysed 137,210 domains in May 2026 and found 97% of published llms.txt files received zero traffic. No AI bot ever requested a file that did not exist. Google’s John Mueller, June 2025: “no AI system currently uses llms.txt.”

Google-Extended, as an AI Overview control

This is the most consequential misunderstanding in the category. Google-Extended governs Gemini training and grounding. Google states it “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal.” AI Overviews are fetched by Googlebot. Blocking Google-Extended does nothing to AI Overviews.

Length

Correlation between word count and AI Overview citation is Spearman 0.04. 53.4% of cited pages are under 1,000 words; 16.6% under 350. Median word count by page type among cited pages: listings 315, core pages 317, articles 1,166.

6. The controls that do exist

Because eligibility requires a page be showable “with a snippet,” the snippet controls are the real on/off switch.

There is no way to appear in Google Search while opting out of AI Overviews, short of suppressing snippets — which also degrades normal search appearance. That is a real trade-off, and publishers should make it deliberately rather than discovering it.

7. What actually to do

The evidence supports a short list. Everything on it follows from a documented finding above.

1. 1

   Target question-form queries and their sub-queries

   Query form is the strongest predictor of Overview presence — 60% and 64.7% for question queries versus 8–10% for short ones. Build coverage for the natural-language questions in your category and for the sub-queries a fan-out would generate, not only the head term.

2. 2

   Put the direct answer immediately under the heading that asks the question

   Retrieval operates on passages. A page that answers in its fourth paragraph is competing against pages that answer in their first.

3. 3

   Keep the answer in text

   Google's own best-practice list names this: keep important content in textual form. Content locked in an image is content that cannot be extracted.

4. 4

   Stop optimising for length

   With a 0.04 correlation between word count and citation, adding words to a page is not an AI Overview tactic. Specificity is.

5. 5

   Build brand demand, not only topic rankings

   Branded queries gained 18.7% CTR where non-branded lost 20.0%, and branded terms rarely trigger Overviews at all. This is the strongest defensive position available.

6. 6

   Do not stake the programme on ranking #1

   The #1 result is cited 43–58% of the time, and a meaningful share of citations come from pages outside the top 10 entirely. Topical coverage of the fan-out beats position on the head term.

7. 7

   Treat YouTube as a citation surface

   5.6% of all AI Overview citations, and 18.2% of non-ranking ones — up 34% over six months.

8. How to measure it without fooling yourself

AI Overviews are unstable. Ahrefs tracked 43,000+ keywords with at least 16 observations each over a month: a 70% chance of changing between consecutive observations, average persistence 2.15 days, and 45.5% of cited sources entirely new between consecutive responses. Yet cosine similarity between versions was 0.95 — the sources rotate, the answer does not.

Presence itself is unstable at platform level too. Semrush’s series on a fixed 10M-keyword set moved 6.5% → 24.6% → 15.7% inside eleven months of 2025.

Five rules follow.

• A single check proves nothing. With 70% inter-observation change, one observation is noise.
• Report citation share across N observations in a defined window, never presence on a date.
• Track presence rate separately from citation rate. Otherwise a platform-wide shift in how often Google shows an Overview reads as your performance changing.
• Search Console cannot isolate this. AI Overview and AI Mode impressions are folded into ordinary Search totals with no filter. External repeated sampling is the only route.
• Check attribution accuracy, not just presence. The Washington University audit found 11.0% of AI Overview claims unsupported by the pages cited, with omission the dominant failure. Being cited for something you did not say is a distinct risk.

Because the substance of the answer is stable while the sources rotate, compete to be a source of the claim the Overview already makes consistently, rather than chasing a specific citation slot.

9. What we could not verify

• Which content formats earn citations. We found no credible primary study measuring citation rates by format — lists versus tables versus direct answers. The topic is dominated by vendor content with no disclosed methodology. Common advice here is untested.
• Trigger volatility for a fixed query set. No dedicated study exists; only indirect evidence from prevalence time series.
• Sample sizes for BrightEdge, Similarweb, Advanced Web Ranking and SparkToro are not disclosed by their publishers. Every figure from those sources is labelled accordingly above.
• Google's own announcement pages for the FAQ deprecation, the May 2025 AI-search guidance, and the June 2026 Search Console generative-AI reports render client-side and could not be retrieved directly. Where we rely on them, we say so.

Sources

Google primary documentation

• Google Search Central. AI features and your website. Updated 10 December 2025.
• Google Search Central. Structured Data General Guidelines. Updated 10 July 2026.
• Google Search Central. Creating helpful, reliable, people-first content. Updated 10 December 2025.
• Google Search Central. Google common crawlers.

Academic

• Chapekis, A., Lieb, A., Shah, S., & Smith, A. (2026). Investigating Click Behaviors On Google Search Result Pages That Produce an AI Overview. arXiv:2608.04831
• Xu, H., Iqbal, U., & Montgomery, J. M. (2026). Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact. Washington University in St. Louis. arXiv:2605.14021

Independent measurement

• Pew Research Center. Google users are less likely to click on links when an AI summary appears in the results. 22 July 2025.
• Ahrefs (Law & Guan). Update: AI Overviews Reduce Clicks by 58%. 4 February 2026.
• Ahrefs (Linehan & Guan). Update: 38% of AI Overview Citations Pull From The Top 10. 2 March 2026.
• Ahrefs (Linehan & Guan). 76% of AI Overview Citations Pull From the Top 10. 21 July 2025. Superseded by the March 2026 update.
• Ahrefs (Linehan & Guan). AI Overviews Change Every 2 Days (But Never Change Their Mind). 11 November 2025.
• Ahrefs (Gavoyannis & Guan). Short vs. Long Content in AI Overviews. 3 December 2025.
• Ahrefs (Linehan & Guan). We Tracked 1,885 Pages Adding Schema. 11 May 2026.
• Ahrefs (Linehan & Guan). We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read. 15 June 2026.
• Semrush. Semrush AI Overviews Study. Updated 15 December 2025.
• seoClarity. The Overlap Between AI Overviews and Organic Rankings. Data 12 October 2025.
• Amsive (Guevara). Google AI Overviews: New CTR Study. 16 April 2025. Date range not disclosed.
• Originality.AI. 52% of AI Overview Citations Appear in the Top-10 Google Search Results. 18 November 2025. Sample not disclosed.
• BrightEdge. AI Overviews at the One-Year Mark. 12 February 2026. Sample not disclosed.
• SparkToro (Fishkin). In 2026, Less than One Third of Google Searches Still Send a Click. 9 June 2026. Panel size not disclosed.
• Search Engine Journal. Google Drops FAQ Rich Results From Search. Secondary reporting of Google's statement.
• Search Engine Land. Google AI Mode traffic data comes to Search Console. 16 June 2025.
• Search Engine Roundtable. Google Says No AI System Currently Uses LLMs.txt. 17 June 2025.