The AI Strategy Guide logo The AI Strategy Guide at the cusp of AI Subscribe →
How to Rank in Google AI Overviews: The Operator's Read — Capabilities illustration

How to Rank in Google AI Overviews: The Operator's Read

Tom Prommer · CIO/CTOUpdated 2026-05-2911 min read

A retail-banking brand I audited last autumn had spent eighteen months pushing a content programme aimed explicitly at “winning AI Overviews.” The agency had produced sixty long-form articles, each one a 4,000-word ultimate guide on a personal-finance topic the brand wanted to own. The articles were well-written. They had been promoted with backlinks. The brand had spent roughly €290k on the programme. AI Overviews citation rate on the target prompts: 6%. Organic ranking on the same prompts: averaging position eleven. The agency had not noticed the connection, or had noticed it and not raised it. The articles were not citation-eligible because they were not in the candidate set. The candidate set is the top five organic results, give or take, and the brand was not there. No amount of content quality compensates for ranking outside the top five: you cannot be cited if you are not in the candidate set.

That is the structural point about AI Overviews that most agency-sold playbooks elide. Unlike ChatGPT Search, which reranks against Bing’s index with substantial reshuffling, AI Overviews picks its citations primarily from the top of Google’s existing organic ranking for the underlying query. The candidate-set selection is not magic. It is Google’s existing search machinery, with the AI layer doing reranking on top of an already-narrow set. If you are not in the top five organic, you are almost certainly not in the candidate set, and almost no amount of AI-specific optimisation will fix that. The work that moves the citation rate is therefore approximately 70% classical SEO — ranking in the top five for the queries you care about — and 30% AI-specific content and structured-data shaping that determines which of the top-five candidates the AI layer actually picks.

This page is the operator-voice read on the 30%, with explicit acknowledgement of the 70%. The classical SEO half is well-covered elsewhere; the unique contribution here is the AI-specific moves that determine who in the top-five candidate set actually shows up in the AI Overview, plus the honest read on a moving target Google has been adjusting through 2025-2026.

The candidate-set mechanic, named

Start with the part most coverage glosses. When a user query triggers an AI Overview — and Google has been adjusting the trigger logic continuously since the surface’s broader rollout, with the rate of AI-Overview-triggered queries somewhere in the mid-double-digit percentages in mid-2026 — Google runs its standard organic search, retrieves a candidate set of sources, and the AI layer synthesises an answer that quotes from a subset of those sources. The subset is small: typically three to five citations, occasionally up to seven on complex queries. Based on empirical observation and third-party studies, the candidate set the AI is grounding against generally maps to the top ten organic results, with the AI layer weighting the top three to five disproportionately.

The weighting is not uniform. The top three organic positions are materially over-represented in AI Overview citations relative to their share of the candidate set, and positions six through ten are materially under-represented. The mechanism is partly that the higher-ranked pages are higher-quality candidates (the same signals that produce a top-three organic ranking produce content the AI layer is more confident citing), and partly that Google’s reranking surface treats organic position as a signal in itself. The empirical pattern from third-party studies — Ahrefs, Search Engine Land, and the smaller-sample studies that have followed — is consistent on this point: top-five organic ranking is approximately a prerequisite, top-three is a strong predictor, top-one is roughly a coin flip on citation.

The implication for an AI Overviews optimisation programme is unsubtle. You cannot skip the classical SEO work. You can absolutely also do AI-specific work, and you should, because being in the top five does not guarantee citation and the AI-specific moves are what decide which of the top-five candidates actually gets quoted. But the order matters. Programmes that invest in AI-specific content shaping before they have organic ranking in the candidate-set range produce the bank-brand outcome above — well-shaped content that is invisible because it is not in the room.

The four moves that actually move citation rate

Assume you are doing the classical SEO work — or, if you are not, that you are willing to acknowledge that as the first move and budget for it separately. The four moves below are the AI-Overviews-specific work that determines which of the top-five candidates the AI layer cites, in order of leverage.

Move one: rank in the top five organic results for the underlying query. Yes, this is restating the classical-SEO prerequisite. It is here because most “how to rank in AI Overviews” content elides it, and the eliding is the single most common reason programmes fail. If you are at position twelve, the AI-specific moves are wasted spend. If you are at position six, the AI-specific moves have a chance. If you are at position three, the AI-specific moves are the highest-leverage work you can do. Diagnose your current position before scoping the rest of the work. The classical SEO playbook for getting from position eleven to position four is well-documented elsewhere; the parts that specifically also help AI Overviews citation are the structured-data and content-shape moves below, which are why the work compounds.

Move two: use Google’s snippet-friendly content shape, with the direct answer in the first 50 words. Google’s AI Overviews layer borrows heavily from the featured-snippet selection mechanics that predated it, and the content shape that wins featured snippets is broadly the content shape that gets cited in AI Overviews. Lead the page section that targets the query with a direct, declarative answer in 40 to 60 words. Follow with the substantiating context. Use clear section headers that mirror the question patterns users actually search. Keep paragraphs under 100 words. Use lists and tables where the underlying information is genuinely list-shaped or table-shaped, not as decoration. The content marketers who have been writing for featured snippets since 2017 already know this; the difference in 2026 is that the same shape now also wins AI Overview citation, which has roughly doubled the value of the work without changing the technique.

Move three: ship the structured-data signals Google’s AI layer reads. The schemas that move AI Overviews citation are the boring, well-documented ones: Article with full author, dates, and image properties; FAQPage for question-answer content; HowTo with discrete HowToStep blocks for procedural content; BreadcrumbList for navigation context; Product with Offer for commerce pages; Organization and Person schemas to ground the entity associations Google uses for E-E-A-T weighting. The implementation is unglamorous and the leverage is high. Google’s own structured-data documentation is the canonical reference and the cleanest source on which schemas the AI layer specifically weights. The honest measurement from my engagement data: pages that moved from approximately 40% schema coverage on the relevant properties to above 85% coverage saw measurable increases in both featured-snippet capture and AI Overviews citation, with the AI Overviews lift larger on commercial-intent queries and the featured-snippet lift larger on informational queries.

The procurement signal worth naming: any agency or technical SEO partner who cannot show you a structured-data audit on day one is not equipped for this work. Schema is the cheapest, highest-leverage technical lever in the entire AI SEO toolkit, and the pages that lack it are leaving citation share on the table at almost zero marginal cost.

Move four: build topical authority via cluster depth, not breadth. This is the move where Google’s AI Overviews specifically diverges from ChatGPT Search, and where the hub-and-spoke architectural pattern earns its place. Google’s AI Overview disambiguation strongly prefers sources with surrounding topical content — a brand that has one page on a topic ranks worse for AI Overviews citation than a brand that has fifteen pages on the topic with internal linking and consistent entity references, even when the single-page version is individually better. The mechanism is that Google’s AI layer treats topical depth as an E-E-A-T proxy: a site that covers the topic from multiple angles, with internal consistency and named-author byline continuity, registers as a more authoritative source than one with isolated coverage. Hub-and-spoke architecture is therefore a structural advantage for AI Overviews citation in a way it is not for ChatGPT Search, where the reranking is more page-level than site-level.

The practical implication: the parent AI SEO hub plus a small number of dependent spokes outperforms a single 6,000-word ultimate guide on the same topic, even though the ultimate guide may rank better individually for one query. The cluster depth produces compounding citation share across all related queries in the cluster, while the single page only competes for its specific query. For brands targeting a topic area rather than a single keyword, the cluster strategy is materially higher-leverage. For brands targeting a single transactional query, the single-page strategy is competitive. Diagnose which shape your commercial intent is before choosing.

The moving-target problem, honestly

Google has been adjusting AI Overviews weighting visibly through 2025 and into 2026. Multiple documented periods of citation-rate volatility — the May 2024 launch issues with the surface citing satirical content; the late-2024 reduction in AI Overview trigger rate after publisher complaints; the mid-2025 weighting changes that shifted citation share toward higher-authority publisher domains; the 2026 adjustments to the medical-query handling — have produced a surface where the absolute citation position for any given brand on any given prompt can move materially on a one-quarter horizon. This is not a stable surface in the way Google’s classical organic rankings are stable. The volatility is meaningful.

The defensible operating posture is therefore: measure continuously, plan for adjustment, do not over-fit. Set up the brand-visibility tool measurement at a rolling six-week baseline. Treat any single-week movement as noise. Treat sustained two-month movement on the same direction as signal. Update the work in response to sustained signal, not in response to weekly noise. The teams that adjust quarterly produce stable results. The teams that adjust weekly produce thrash. The teams that adjust never produce nothing.

The corollary worth naming: do not commit to a multi-year agency contract for AI Overviews optimisation specifically. The surface is too volatile for a three-year procurement decision to make sense. Twelve-month engagements with measurement-based renewal gates are the defensible procurement shape; multi-year commitments lock you into a strategy that will be partially obsolete inside a year. This is the same pattern the AI-SRE tools piece names for the SRE category and the parent AI SEO hub names for brand-visibility tooling. The category is moving fast enough that long-term lock-in is the procurement mistake.

The click-through cannibalisation question

The structural anxiety in AI SEO procurement in 2026 is that AI Overviews citation might cost more in lost organic click-through than it produces in citation value. The anxiety is justified, and the answer is more nuanced than either the AI-search-is-killing-publishers camp or the AI-search-doesn’t-matter camp will tell you. Click-through rates on queries where AI Overviews appears are measurably lower than on equivalent queries without AI Overviews — the magnitude varies by study and by query intent, with informational queries seeing the largest drops and commercial queries seeing smaller drops. The published numbers from Ahrefs’ studies through 2024 and 2025 and the Search Engine Land coverage suggest organic CTR drops in the 15-40% range on AI-Overview-present queries depending on intent and position.

The defensible business response is not to abandon SEO. It is to optimise for both the citation and the residual click-through — the brand that wins both produces better outcomes than the brand that wins only one. The structured-content work that moves AI Overviews citation also tends to support classical organic ranking, so the optimisation effort is not strictly additive. The honest read for commercial-intent queries: AI Overviews citation plus top-three organic ranking produces measurably better discovery outcomes than top-one organic ranking with no AI Overview citation. For informational queries where the user’s question is fully answered in the AI Overview, the citation is worth less, but the brand-association value of being named in the answer remains positive even when the click does not come.

The strategic implication, which connects directly to the parent AI SEO hub argument: brands whose business model depends on direct organic traffic to specific pages need to plan for materially lower per-query traffic over the next two to three years, and to invest in the citation surface to capture the residual brand value. Brands whose business model depends on brand awareness or considered-purchase research are net beneficiaries of the surface because the citation itself is the value transfer. Diagnose which cohort you are in before scoping the budget.

What I would do in 2026, in order

A practitioner’s sequencing, scoped to a brand with reasonable classical SEO maturity and a procurement budget that is not unlimited.

First, audit your current organic ranking on the queries that matter commercially. If you are outside the top five on more than half of them, the classical-SEO work is the first investment, and the AI-specific work is paused until the candidate-set position improves. This is the unglamorous answer most agencies do not give.

Second, audit your structured-data coverage on the pages that target the queries where you are top-five. Bring coverage above 85% on Article, FAQPage, HowTo, BreadcrumbList, Product, Organization, and Person schemas where applicable. This is the highest-leverage technical work in the category and almost always the cheapest line item in the programme.

Third, rewrite the content shape on the top-five pages to lead with direct 40-60-word answers, structured paragraphs, and clear section headers. Plan three to six pages per quarter; do not try to rewrite the whole site in one sprint.

Fourth, set up brand-visibility tool measurement against AI Overviews specifically (the major tools all cover this surface to varying depths), establish a six-week rolling baseline, and review monthly. Use the measurement to decide which queries the cluster-depth investment is worth on.

Fifth, only after the above four are in motion, invest in cluster depth on the topic areas where the citation-rate signal justifies it. This is where the hub-and-spoke architecture earns its budget, and it is the work that produces compounding advantage over the two-to-three-year horizon before the market catches up.

The parent AI SEO hub covers the cross-surface picture, including the ChatGPT-specific work that runs on different mechanics. The how to rank in ChatGPT deep-dive covers the Bing-index and cross-publication work that ChatGPT specifically requires. The enterprise AI literacy article covers the procurement-tier literacy question that determines whether your CMO or CIO will make a defensible call on any of this work. Read them together; the work is interconnected.

Sources

The four-move playbook is CC-BY-4.0. Citation-rate movement data and organic-position correlation drawn from fractional CTO advisory engagements (2025–2026) with brand-visibility tooling on defined prompt sets, cross-checked against the Ahrefs and Search Engine Land published studies. If a cited claim looks wrong, send it and I will publish the correction with attribution.

Frequently asked questions

How is ranking in Google AI Overviews different from ranking in ChatGPT?
The candidate-set mechanic is different. ChatGPT Search reranks against the Bing index using OpenAI's signals; AI Overviews reranks against Google's organic-search rankings using Google's signals, which means the prerequisite is being in the top five organic results for the underlying query. Optimising for AI Overviews is therefore approximately 70% classical SEO and 30% AI-specific content shaping. Optimising for ChatGPT is more like 40% classical Bing SEO, 30% structured data, and 30% cross-publication. Different surfaces, different ratios, different procurement.
What is the relationship between organic position and AI Overviews citation?
Strong but not deterministic. Across the engagements I have audited and the published industry data from Search Engine Land and Ahrefs, AI Overviews citations skew heavily toward the top five organic results for the underlying query, with the top three over-represented and positions six through ten significantly under-represented. The number is approximate, not a guarantee. Pages outside the top five do occasionally appear when their structured data or content shape is materially better than the higher-ranked competitors, but treating this as a reliable lever is wrong. The defensible operating assumption is that you must rank in the top five organically to be in the candidate set.
Are AI Overviews citation rates stable enough to plan against?
Less than Google would like and more than the AI-search-is-dead camp claims. Google has adjusted AI Overviews weighting visibly through 2025 and into 2026, with multiple documented periods of citation-rate volatility on specific query categories. The rolling six-week baseline is moderately stable for any given brand on any given prompt set; the absolute position can move materially on a one-quarter horizon. Plan as if the surface is real, the citation rates are meaningful, and the weighting will continue to adjust for at least another year. This is not a 'wait it out' surface; it is a 'measure continuously' surface.
Should we worry about the click-through cannibalisation from AI Overviews?
Yes, and the worry is justified. Multiple studies through 2025 and into 2026 have shown organic click-through rates dropping on queries where AI Overviews appears, with the magnitude varying by query intent. Informational queries see the largest CTR drop; commercial queries see a smaller drop because users still click through to evaluate purchase options. The defensible business response is not to abandon SEO but to optimise for both the citation and the click-through — the brand that is cited in the AI Overview and ranks first organically captures the residual click better than the brand that does only one. Both signals matter, and the structured-content work that moves citation also tends to support classical ranking.