How to show up in Google AI Overviews (and why you might not be)

If your site is missing from AI Overviews, it is missing from a results page where most searches now end without a click to any website: according to SparkToro, 68% of Google searches in 2026 ended without a click. The scale is not marginal either. A Semrush study of 10 million keywords found that 13.14% of all Google searches triggered an AI Overview in March 2025.

The first thing to correct is the premise. There is no separate "AI Overviews algorithm" you can target with a trick, a tag or a payment. Google's own Search documentation is consistent on this: AI Overviews draw on the same systems as ordinary search, standard SEO guidance applies, and there is no AI-specific markup. Google states plainly that structured data is not a requirement for appearing, and that files like llms.txt are ignored by Google Search. What you can do is twofold: earn the organic rankings that feed AI Overviews, and structure your pages so they are the easiest thing on the results page to quote.

AI Overviews are one surface of what AI visibility is, whether AI engines mention and recommend your brand when buyers ask. This guide covers how Google picks its AI Overview sources, the tactics that correlate with being cited, and the six reasons sites fail to appear, with a fix for each.

How do AI Overviews choose their sources?

AI Overviews are generated by Google's Gemini models grounded in live search results. Grounding means the model is given real, freshly retrieved web pages to answer from rather than relying on its training data alone: Google retrieves ranking pages for the query and its related sub-questions, composes an answer, and cites the pages it drew from. Ranking and AI Overview citation therefore overlap heavily. A Rich Sanger / Authoritas study found that a link in organic position 1 had a 53% chance of appearing in the AI Overview for that query, versus 36.9% at position 10.

Two things follow from that mechanism, and they shape every tactic below.

First, the query you see is not the only query that matters. Google "fans out" a search into related sub-questions and retrieves sources for each. A Search Engine Land study of 10,000 keywords found that pages ranking for these fan-out sub-queries are 161% more likely to be cited, and that fan-out accounts for 51% of all AI citations. A page that answers "best accounting software for a UK freelancer who invoices in euros" can be cited in the AI Overview for "best accounting software", because it wins the sub-question, not the head term.

Second, ranking gets you retrieved; extractability gets you cited. Extractability is how easily a passage can be lifted out of your page and used as a complete answer on its own. Once Google has a set of ranking pages, it quotes the passages that answer most directly. Structure (direct answers, question headings, sourced statistics) is what separates the cited page from the merely retrieved one.

It also matters whose pages tend to win. According to a Surfer SEO study, the five most-cited domains in AI Overviews are youtube.com (23.29%), wikipedia.org (18.41%), google.com (16.38%), reddit.com (9.37%) and linkedin.com (8.80%). That mix shifts sharply by industry. Surfer SEO and Amsive data shows that in health, nih.gov leads at 38.89%, with youtube.com at 27.91% and healthline.com at 15.06%. The practical point: your own site is only one of the surfaces Google can cite you from, and the surfaces that matter are specific to your category. This is the same dynamic covered in how AI assistants decide which brands to recommend.

What triggers an AI Overview?

Not every search produces an AI Overview, and the pattern is predictable: longer, informational, question-shaped queries trigger them most. An SEO.com AI Overview keyword study found long-tail queries of six to eight words have a 32–57% chance of showing an AI Overview; informational queries trigger one 39% of the time; and informational queries of eight or more words trigger one 67% of the time.

This changes what you optimise for. Short head terms are the queries least likely to show an AI Overview at all. The conversational, multi-constraint questions real buyers ask ("best accounting software for a UK freelancer who invoices in euros", not "accounting software UK") are the queries where AI Overviews appear most often, and where the citation is winnable. If your content plan still targets two-word keyword strings, you are targeting the queries least likely to produce an AI Overview in the first place.

Can you optimise for AI Overviews?

Yes, but "AI Overviews optimisation" means stacking the probabilities, not flipping a switch. The seven tactics below are the ones that correlate with inclusion in the published research. None is a guarantee: AI answers are probabilistic and vary between runs, and correlation in these studies is not proof that any single tactic caused the citation.

One caveat worth stating before the list. Much of the published extraction research measures citations in ChatGPT rather than in Google AI Overviews specifically. The patterns are consistent enough across studies to act on, and they align with what Google says about ordinary search, but treat cross-engine findings as indicative rather than measured for AI Overviews.

1. Rank for the question, and for its sub-questions. Ranking and citation overlap heavily: 40–76% of AI Overview citations also appear in the top 10 organic results (Authoritas and Optimizely studies). Build pages around the specific questions inside your topic, not just the head term: fan-out sub-queries drive 51% of AI citations, per Search Engine Land's 10,000-keyword study.

2. Phrase headings as the questions people type. A Kevin Indig analysis of 1.2 million ChatGPT citations found question marks in headings roughly double citation rate, 18% versus 8.9%. Question headings also map one-to-one onto the fan-out sub-queries Google retrieves for.

3. Answer directly in the first sentence under each heading. Per AmiCited data, a direct answer in the opening paragraph raises retrieval frequency by around 40%. Front-load ruthlessly: Kevin Indig's analysis of 1.2 million ChatGPT citations found 44.2% of citations come from the first 30% of a page.

4. Write in self-contained chunks. Ekamoira research found self-contained chunks of 50–150 words earn 2.3 times more AI citations. Each paragraph should carry one idea and survive being quoted with nothing around it, no "as mentioned above", no pronouns pointing at the previous section.

Here is the difference in practice. Both passages below are invented examples of writing style, not findings.

Hard to extract: "As we discussed earlier, there are several factors to consider here, and the right choice really depends on your situation. Many of the options mentioned above could work well."

Easy to extract: "For a UK freelancer who invoices in euros, the deciding factors are multi-currency invoicing, HMRC-recognised Making Tax Digital support, and pricing under £20 a month. Three of the eight tools tested meet all three."

The second version names its entities, answers a specific question, and works as a complete quote. That is the shape AI Overviews lift.

5. Include sourced statistics. The Princeton and Georgia Tech paper on generative engine optimisation (GEO, the practice of shaping content so AI engines quote it), presented at ACM KDD 2024, found that adding statistics lifts AI-answer visibility by 15–40%, and quotations by 30–40%. Weld the source into the sentence, as this article does, so the attribution travels with the quote.

6. Keep content genuinely fresh. Ahrefs' 17-million-citation study across seven AI platforms found AI assistants prefer content around 25.7% fresher than what organic Google returns, and an AirOps study found pages with a visible "last updated" timestamp receive 1.8 times more citations. Refresh the substance (numbers, examples, findings), not just the date.

7. Build presence on the sources Google already cites in your category. If AI Overviews in your niche lean on YouTube, Reddit or specific review sites (per the Surfer SEO domain data above), being named on those pages is a second way into the answer that does not depend on your own rankings. A Search Engine Land study of 800+ websites across 11 industries found multi-platform brand mentions correlate with AI visibility at r=0.87, a strong association, though the study measures correlation, not cause. Which sources matter for your category is a measurement question. More on that below, and in where AI engines get their answers.

Why is my website not showing in AI Overviews?

A website usually misses AI Overviews for one of six reasons: it does not rank in the top 10, its target queries rarely trigger an AI Overview, its content is hard to extract, its content is stale, it is absent from the third-party sources Google cites in that category, or its pages are blocked from being quoted. Work through them in that order. They run from most to least common.

1. You don't rank, for the query or its sub-questions. This is the most common cause and the least welcome answer: AI Overviews are built from search results, so a site with no top-10 presence starts at a real disadvantage. It is not a closed door (between a quarter and three-fifths of AI Overview citations sit outside the top 10, on the Authoritas and Optimizely numbers above), but the odds are much better inside it. The fix is orthodox SEO, aimed at the fan-out layer: the specific sub-questions inside your topic are less competitive than head terms, and pages ranking for them are 161% more likely to be cited (Search Engine Land, 10,000 keywords).

2. The queries you care about don't trigger AI Overviews. Before diagnosing your content, check whether an AI Overview even appears for your target queries. If you sell against short commercial terms, there may be no AI Overview to appear in, and your real question is about Gemini, ChatGPT and Claude instead, where the same buyer questions get answered conversationally. That is AI visibility in the broader sense: whether AI engines mention and recommend you when buyers ask.

3. You rank, but you don't get cited. This is the extractability gap: Google retrieves the page but quotes a competitor whose answer is easier to lift. Rework key pages with the structure from the tactics section: direct first-sentence answers, question headings, sourced numbers, atomic paragraphs. Among pages with similar rankings, structure is the tiebreaker.

4. Your content has aged out. If a page was written in 2023 and untouched since, the freshness findings above are working against it. Schedule real refreshes: updated statistics, current examples, and a visible, honest "last updated" date.

5. The AI Overview cites your category's ecosystem, and you're not in it. In many categories, AI Overviews cite review sites, YouTube videos, Reddit threads and industry publications more readily than supplier sites. If those citations name competitors and not you, the fix is earned coverage on those specific sources, which requires knowing which ones they are.

6. Technical eligibility. AI Overviews can only cite pages Google can index and is allowed to quote. Check that key pages are indexed, and that snippet controls (nosnippet, data-nosnippet, restrictive max-snippet values) aren't stopping Google from quoting the content. This is rarely the cause, but it is the cheapest to rule out.

How do you know which of these applies to you?

You measure, rather than guess. This is the step Skavora exists for: the free check researches your brand, generates 25 realistic buyer questions, and runs them live through Google's Gemini with web grounding. Around 80% of those questions are non-branded (they never name your brand) because the questions that win new customers are the ones where the buyer does not yet know who to ask for. You see whether you are named, who is named instead, and which sources the answers cite, ranked by influence. The £7.99 report runs the same questions across four engines (Gemini, ChatGPT, Claude and Google AI Overviews) so you can compare which sources AI Overviews cite in your category against the other three. Run a free AI visibility check to see which of the six problems above is yours.

Two honest caveats. A scan is a snapshot: live AI answers vary between runs, and Semrush's AI Visibility Index, tracking 2,500 prompts across Google AI Mode and ChatGPT, found 40–60% of AI-cited sources change from month to month. And a scan measures; it does not control what any engine says. What it tells you is where you stand today and where to direct effort next, which is an argument for re-scanning on a cadence rather than a reason to distrust a single measurement.

The eight checks that decide whether you appear in AI Overviews

• Rank first: 40–76% of AI Overview citations also appear in the top 10 organic results (Authoritas/Optimizely). Target fan-out sub-questions, not just head terms.
• Phrase headings as questions; question marks in headings roughly double citation rate (Kevin Indig, 1.2M citations).
• Answer completely in the first sentence beneath each heading: a direct opening answer raises retrieval frequency ~40% (AmiCited).
• Write 50–150-word self-contained chunks (Ekamoira: 2.3x more citations) with sourced statistics welded into the sentence.
• Refresh substance regularly and show an honest updated date (AirOps: 1.8x more citations with a visible timestamp).
• Earn presence on the third-party sources AI Overviews already cite in your category: review sites, YouTube, Reddit, industry publications.
• Rule out snippet controls and indexing issues.
• Measure: run the same buyer questions through the engines and see who actually gets cited.

Skavora measures AI visibility; it does not control AI outputs. All statistics above are attributed to their original studies. Sources: ahrefs.com, searchengineland.com, semrush.com, seo.com, ziptie.dev.