How to Show Up in AI Overviews: A 2026 Playbook
Google’s own documentation answers this question in one sentence, and the answer is not the one being sold: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
That is a direct quote from Google Search Central’s page on AI features. The same page adds that you do not need to create machine-readable files or AI text files, and that there is no special schema.org structured data required. If you have been quoted a fee for an AI Overviews optimisation package, that quote is the first thing worth reading.
So this guide does not pretend there is a secret lever. It does something more useful: it explains the two gates a page has to pass, shows what the controlled data says about which pages actually get picked (the answer will annoy anyone who assumes their number-one ranking is safe), and gives you a sequence of things that genuinely change your odds. It also tells you which parts of the popular advice have no traceable evidence behind them at all.
The two gates, and why the difference matters
Almost every confused conversation about AI Overviews comes from collapsing two separate things into one. There is an eligibility gate, which is technical and documented, and there is a selection step, which is not documented and behaves unlike classic ranking.
Google states the eligibility gate plainly: “To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet. There are no additional technical requirements.”
Read that carefully, because it contains the only genuinely new technical failure mode in this whole topic. A page that is indexed but carries a nosnippet directive, or an aggressive max-snippet limit, is a page Google has been told not to quote — and a supporting link in an AI Overview is a form of quoting. Publishers who added snippet restrictions during the 2024 scraping panic quietly removed themselves from these surfaces. Very few of them know it.
| Control | What it does | Effect on AI Overviews and AI Mode |
|---|---|---|
noindex |
Keeps the page out of the index entirely. | Cannot appear. No index, no eligibility. |
nosnippet |
Tells Google not to show any text snippet for the page. | Removes snippet eligibility, which is a stated requirement. |
max-snippet:[n] |
Caps the snippet at n characters. | A very low cap restricts what Google can surface from the page. |
data-nosnippet |
Excludes a specific block of HTML from snippets. | Useful surgical tool — fence off boilerplate, not your answer. |
Google names exactly these four as the way “to limit the information shown from your pages in Search”. They work as designed. The mistake is applying them broadly and then wondering why the AI never cites you.
The second gate — which of the eligible pages actually get chosen — is where the interesting data lives.
Which queries even trigger one? Nobody agrees, and that matters
Before optimising for a surface, it is worth knowing how often it appears. Here the published figures diverge so wildly that quoting any single one is misleading.
| Source | Reported trigger rate | What it actually measured |
|---|---|---|
| Semrush sensor data, 2025 | About 16% of tracked queries by November 2025, after peaking near 25% in July | A tracked keyword set, US, moving through a year of heavy Google experimentation |
| BrightEdge, 2026 | Around 48% of queries | A different tracked set with a different sampling frame |
| Whitespark, May 2025 | 68% of local business queries | 540 manual queries, 3 US cities, 6 industries — a deliberately narrow local slice |
| Semrush keyword study, Sept 2024 | 39.4% of informational searches | Intent-segmented, so not comparable to an all-query rate |
These are not four measurements of the same thing. They use different keyword sets, different countries, different date ranges and different definitions of “a query”. A 16% figure and a 68% figure can both be honest. The practical conclusion is that no industry average tells you anything about your own queries, and the only trigger rate worth acting on is the one you measure across the keywords you actually care about.
What the data does agree on is the shape of the demand. Semrush’s keyword study found 80% of desktop AI Overviews appeared on informational keywords, with transactional keywords under 3% and navigational under 2%. Between 76% and 82% appeared on keywords with fewer than 1,000 monthly searches. Whitespark’s local study found the same pattern inverted for local intent: a query like “plumbers in Phoenix” showed a local pack 93% of the time and an AI Overview only 15% of the time, while an informational query in the same industry showed an AI Overview 92% of the time.
So the honest summary for a business owner: AI Overviews have taken the long-tail, low-volume, question-shaped, informational query. They have largely not taken the query where someone is ready to buy from you. That distinction should drive where you spend.
The finding that should change your assumptions
The single most useful number in the Semrush study is this: only 20% to 26% of the links inside an AI Overview matched the top ten organic results for the same query. The number-one organic result appeared in just 46% of desktop AI Overviews and 34% of mobile ones.
Read that as both a warning and an opening. The warning: ranking first does not guarantee inclusion, so a page you thought was untouchable is not. The opening: three quarters of the citation slots are going to pages that are not in the classic top ten. If you are stuck at position 12 for a valuable question, the AI Overview is arguably an easier target than the blue links above you.
Two other structural facts from the same study are worth holding on to. The average AI Overview ran 119 words on desktop and 91 on mobile, and carried roughly eleven links. Eleven citation slots against ten organic positions is not a contraction of opportunity — it is a redistribution of it, towards pages that answer one specific sub-question unusually well.
Query fan-out: you are being matched to questions nobody typed
Google’s documentation confirms that AI Overviews and AI Mode “may use a ‘query fan-out’ technique”, issuing multiple related searches across subtopics and data sources, which lets Google “display a wider and more diverse set of helpful links associated with the response than with a classic web search”.
That single mechanic explains the 20-26% overlap. Your page is not competing for the query the user typed. It is competing for one of several sub-queries Google generated from it — questions the user never asked and may not have thought of.
The practical consequence is a change in how you structure a page. A page built as one long argument towards a conclusion is hard to fan-out against. A page built as a set of clearly-labelled, self-contained answers gives Google many small targets. This is not a new technique; it is the same passage-level thinking that has driven good on-page SEO for years, now with a higher payoff.
A worked example. A Singapore accounting firm has one page called “Corporate tax services”. Under fan-out, it competes badly, because the underlying questions — what the corporate tax rate is, when Form C-S is due, what the partial exemption is worth, whether a dormant company still files — are each a separate sub-query. Split into clearly-headed sections with a direct answer in the first two sentences of each, the same content becomes eligible for several fan-out branches instead of none.
What actually improves your odds, in order
1. Audit eligibility before anything else
Check that the pages you care about are indexed, return a normal snippet in a live search, and carry no nosnippet or restrictive max-snippet directive. Check your robots.txt has not accidentally blocked Googlebot from resources the page needs to render. This costs an afternoon and is the only step Google explicitly requires. Our technical SEO basics guide covers the mechanics.
2. Pick the queries where the prize is real
Run your target keyword set and record, per keyword, whether an AI Overview appears at all and whether it appears above or below the fold. Rank the list by commercial value, not volume. A question that triggers an AI Overview and ends the session is a poor investment; a question that triggers one and still leads to a quote request is a good one. Proper keyword research now has this extra column in it.
3. Restructure for passages, not pages
Give each sub-question its own heading. Put the direct answer in the first sentence or two under that heading, then the nuance. Keep the answer self-contained enough to make sense lifted out of context, because that is precisely what happens to it. Use tables for anything comparative — they are dense, unambiguous and easy to extract.
4. Go multimodal, and update the listings you forgot about
Google’s guidance for AI experiences is explicit that you should support text with high-quality images and video, and — the part almost everyone skips — ensure your Merchant Center and Google Business Profile listings are current. Those feeds are inputs to AI experiences, not just to Shopping and Maps. If you sell products or serve a location, a neglected Google Business Profile is a hole in your AI visibility, not merely your local one.
5. Publish something a model cannot synthesise
The strongest experimental evidence here comes from the KDD 2024 paper that coined the term GEO, by Aggarwal and colleagues. Across a benchmark of roughly 10,000 queries they found visibility gains of 22% to 41% from content changes, with the highest-performing tactics being what they called epistemic authority signals: adding statistics, citing sources, and quoting credible authorities, worth up to about 40%.
The caveat matters and is usually dropped: that study evaluated Google’s top five sources synthesised by GPT-3.5-turbo, which is not the 2026 production stack. Treat it as directional evidence that original, attributable, source-citing content is favoured — not as a specification. The direction happens to align with Google’s own advice to make “unique, non-commodity content”.
6. Do not buy schema as an AI tactic
Google says there is no special structured data needed. Ahrefs tested it: 1,885 pages that added JSON-LD between August 2025 and March 2026, matched against roughly 4,000 control pages, showed no citation lift on AI Overviews, AI Mode or ChatGPT — and a small but statistically significant 4.6% decline on AI Overviews. Implement schema markup for rich results and machine-readable clarity, which are perfectly good reasons. Do not pay a premium for it as a citation lever.
The claims we deliberately left out
A great deal of the advice circulating on this topic cites numbers with no traceable methodology. We checked several and excluded them. You will see it asserted that pages with FAQ schema are weighted “approximately 40% higher” in AI source selection, that structured pages get “roughly 3x more” citations, and that a fixed percentage of citations comes from the first third of a page. None of these has a published sample, date, or method attached. They may be true. There is currently no way to know, and repeating them as fact is how the market got into this state.
The same applies to the widely-quoted Gartner prediction that traditional search volume would fall 25% by 2026. It did not happen: Alphabet’s 2026 results describe search queries at an all-time high, with AI features credited for growth in total and commercial queries. When a forecast anchors a sales deck and the forecast has already been falsified, that is worth saying out loud.
What the click data really says
Pew Research Center gives the most methodologically transparent picture available. Working with browsing data from about 900 US adults through its KnowledgePanel Digital panel, Pew examined 68,879 unique Google searches from March 2025, of which 12,593 triggered an AI summary.
Where an AI summary appeared, the click rate on a standard result fell from 15% of visits to 8%. Links inside the AI summary itself were clicked in about 1% of visits. And sessions ended after the search in 26% of visits with a summary, against 16% without.
Three things follow. First, this is a US sample, not Singapore. Second, the halving is real but the baseline was already low — most searches never produced a click. Third, and most usefully: if only about 1% of visits produce a click on an AI Overview link, then appearing in one is mostly a brand-visibility outcome, not a traffic outcome. Value it that way, budget for it that way, and stop expecting a session-count increase that the mechanics do not support.
The Singapore read
Three local facts shape the priority order here, and they point in a different direction from most of the American commentary.
Google handles 92.46% of Singapore search as of July 2026, according to StatCounter, with Bing a distant 3.37%. Whatever happens in the chatbot market, the surface that decides whether a Singapore business is seen is still a Google results page. AI Mode has now reached over 200 countries and around 100 languages and is available here, but Google has been explicit that it is not the default search experience.
At the same time, Stanford HAI’s 2026 AI Index puts Singapore consumer generative-AI adoption near the top globally at roughly 61%, against about 53% worldwide. Singaporeans are unusually willing to ask an assistant. That makes the chatbot surface worth monitoring, but it does not yet make it the surface where most of your demand sits.
Practically, for a Singapore SME: treat AI Overviews as an SEO outcome inside your existing programme, keep your Google Business Profile and any Merchant Center feed genuinely current, and resist the urge to fund a separate line item. Our view on what that costs and what it should include sits in our guide to SEO pricing in Singapore, and the wider strategic picture is in the AI SEO guide for Singapore.
How to measure it without fooling yourself
Three habits, in increasing order of effort.
Track impressions and clicks separately in Search Console. Google does not break out AI Overview appearances, but the signature is recognisable: a query group whose impressions hold steady or grow while its click-through rate falls. That pattern on informational queries is the most reliable proxy available.
Keep a manual log. Take your twenty most valuable questions, run them monthly, and record whether an AI Overview appeared and whether you were cited. Twenty minutes a month, and it beats every automated visibility score currently on the market for accuracy on your own terms.
Separate AI referrals from organic in GA4. The AI Assistants channel captures chatbot traffic where the medium matches ai-assistant and the referrer is on Google’s list. Crucially, traffic from AI Overviews and AI Mode is excluded from it and classified as Organic Search, so that channel will never show you AI Overview performance. The reports worth watching are covered in our GA4 reports guide, and if you are also tracking paid channels, our conversion tracking guide covers the plumbing.
Frequently asked questions
Can I pay Google to appear in an AI Overview?
No. Supporting links in AI Overviews are earned through organic eligibility and selection, not bought. Google does place ads above and below AI Overviews and inside AI Mode, but those are labelled ads and are a separate purchase from the citation links in the answer itself. Any offer to guarantee an AI Overview placement is selling something it does not control.
Does ranking number one guarantee I appear in the AI Overview?
No, and this is the most commonly held wrong assumption. Semrush’s study of 200,000 US keywords found the number-one organic result appeared in only 46% of desktop AI Overviews and 34% of mobile ones, and that just 20% to 26% of AI Overview links matched the top ten organic results at all. The link set is assembled against fanned-out sub-questions, not copied from the rankings.
Should I block Google from using my content in AI Overviews?
You can, but understand the trade. The controls Google names — nosnippet, data-nosnippet, max-snippet and noindex — are the same controls that govern ordinary search snippets, so restricting them costs you visibility in classic results too. There is no setting that keeps you in normal Search while excluding you from AI Overviews. For most businesses that trade is a bad one.
How long does it take to see a change after restructuring a page?
The same timeframe as any SEO change, because it is the same crawling and indexing pipeline: typically a few weeks for the page to be recrawled and reassessed, longer for competitive queries. There is no separate AI index that updates faster. Our guide to how long SEO takes in Singapore applies here without modification.
Are AI Overviews reducing traffic for Singapore businesses specifically?
There is no published Singapore-specific click study, so anyone quoting one for this market is extrapolating. The Pew figures — organic clicks falling from 15% to 8% of visits when a summary appears — are from a US sample. What is reasonable to expect locally is erosion concentrated on informational and definitional queries, and much less movement on transactional, branded and local ones, since those trigger AI Overviews far less often.
Is it worth writing FAQ sections purely to get cited?
Writing clear question-and-answer sections is worth it because fan-out matches sub-questions and a well-headed answer is easy to extract. Adding FAQPage schema on top of that is a different claim, and the evidence for it as a citation lever is weak — Ahrefs’ controlled test found no lift from adding structured data. Write the answers. Do not assume the markup is what does the work.
Where this leaves you
The uncomfortable truth is that there is no AI Overviews project to run. There is an eligibility check most sites pass and a few quietly fail, a content structure that suits fan-out better than the one you probably have, and a measurement discipline that stops you attributing normal seasonality to the machines.
The genuinely strategic decision is one of allocation. Your informational content is now competing against a summary that answers the question without a click; your transactional, local and branded content is very largely not. Move investment accordingly, and stop paying for the informational click you are no longer going to get.
If you want that triage done properly across your site, our AI SEO service in Singapore runs it inside an SEO engagement rather than as a separate product. Related reading in this cluster: GEO vs AEO vs LLM SEO for the terminology, how to rank on ChatGPT for the chatbot surfaces, and AI SEO tools in 2026 for what the monitoring actually costs. What we have delivered for other Singapore businesses is in our case studies.
Last updated 4 August 2026. Written by Adrian Tan and the SDM team. Sources: Google Search Central, “AI features and your website” (developers.google.com); Google Search Central blog, “Top ways to ensure your content performs well in Google’s AI experiences on Search” (May 2025); Semrush AI Overviews study, 200,000 US keywords, September 2024; Whitespark, “The prevalence of AI Overviews in local search”, 540 queries, May 2025; Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results” (July 2025); Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024; Ahrefs controlled study of schema markup and AI citations (2026); StatCounter Singapore search engine market share, July 2026; Stanford HAI 2026 AI Index.
Related guides
AI Overviews are one surface of five. SEO for AI search sets out the single work programme that serves all of them, in the order the layers actually depend on each other. And if you are trying to decide what to fund over the next two years, the future of SEO and AI works through which forecasts survived contact with the data.
For the background on the surface itself — how often AI Overviews appear, what the Pew click data actually shows, and what it means in a market where Google holds more than 92% of searches — see AI Overviews in Singapore.
For the mechanism underneath all of this — how query fan-out turns one question into many concurrent searches, and why that decoupled citation from ranking — see our guide to Google AI Mode and SEO.
