How do I rank in Google AI Overviews?

Of all the AI surfaces, Google’s AI Overviews are the most tightly coupled to classic SEO. That is good news: your existing work already counts, and the levers are more familiar than they are for the chat assistants.

How AI Overviews pick sources

An AI Overview is generated from Google’s own understanding of the web. Google retrieves relevant pages from its index — the same index that produces the blue links — and generates a summary, citing a handful of sources alongside it.

That has two immediate consequences. First, your organic performance largely determines whether you are eligible, because you have to be in the retrieved set to be quoted from. Second, Googlebot renders JavaScript, so the rendering problem that makes sites invisible to ChatGPT does not apply here in the same way.

AI Overviews behave far more like an expanded featured snippet than like ChatGPT. Optimise accordingly.

Organic strength gates everything

Cited sources are usually drawn from pages already performing well for the query or for closely related ones. Google occasionally pulls a page from further down when it contains a notably clean answer to a specific sub-question, but strong organic performance remains the most reliable route in.

Practically, this means the work is the work you already know. Be indexed. Be relevant. Be internally linked. Have enough authority for the topic. If a page is not ranking in the top ten or twenty for its query, optimising it for AI Overviews specifically is premature — the constraint is upstream.

This is also why the “SEO is dead” framing has been so unhelpful. On the surface that appears on roughly a third of all monitored queries, SEO is the entire entry requirement.

Passage-level optimisation

Once eligible, what decides inclusion is the quality of a specific passage rather than the page as a whole.

Answer the sub-question directly. An AI Overview is assembled from several sources, each contributing a piece. Your best chance is owning one piece cleanly — a two or three sentence answer to one specific sub-question, complete on its own.

Question-led headings. Structure the page so each heading is a real question and the sentences immediately beneath it answer that question and nothing else. This creates clean, extractable blocks.

Front-load the answer. Preamble is fatal here. If the answer arrives in paragraph four, the passage that gets extracted will be someone else’s paragraph one.

Be specific. Numbers, dates, named sources. Generic phrasing gets paraphrased without attribution; a concrete claim gets quoted with a link.

Use lists and tables where the content genuinely is structured. Steps, criteria and comparisons are disproportionately likely to be lifted, because they map cleanly onto how a summary is assembled.

Add matching structured data. Article, FAQPage and HowTo where genuinely applicable, always mirroring the visible text.

Which queries trigger them

AI Overview presence has been remarkably stable through 2026 — around 32.5% of monitored queries, fluctuating only between roughly 31.0% and 34.4%. That narrow band indicates settled deployment rather than continued expansion.

Distribution is uneven in predictable ways. Informational queries — definitions, how-tos, comparisons, “what is”, “how do I” — trigger them far more often. Transactional and navigational queries trigger them much less. Sensitive topics — medical specifics, legal advice, elections, some financial subjects — are frequently suppressed entirely, and Google tunes those thresholds continuously.

That distribution should shape where you invest. If your traffic is predominantly transactional, AI Overviews are a smaller problem and a smaller opportunity than the industry discussion implies. If it is informational, this is where a third of your visibility is being decided.

How this differs from ChatGPT

The practical implication is that these are two different programmes sharing a foundation. A page can be cited in AI Overviews and completely invisible to ChatGPT because it renders client-side. Conversely, a brand can be named constantly by ChatGPT — on the strength of third-party description — while never appearing in an AI Overview because its own pages do not rank.

Our guide to optimising per engine covers running both without duplicating effort.

Is the citation even worth it?

Worth asking honestly, because the answer is nuanced.

Pew Research, tracking roughly 68,000 real queries, found users clicked a result in 8% of visits where an AI summary appeared, against 15% where one did not. Ahrefs measured a 58% click-through decline for top-ranking pages in February 2026, up from 34.5% in April 2025. The click pool is genuinely smaller.

But the comparison that matters is not “cited in an AI Overview” versus “position one before AI Overviews existed”. It is “cited” versus “not cited, on the same page, today”. On that comparison it is clearly worth having: you get the residual clicks, the brand exposure to everyone who reads the summary without clicking, and the implicit endorsement of being the source Google chose.

There is also a quieter benefit. A segment of users has become sceptical of AI summaries after high-profile errors and deliberately scrolls past them. Being the cited source under a summary those users distrust is worth more than it looks, because the sceptical clicker is a high-intent clicker.

Three common errors

Optimising pages that do not rank. Eligibility is gated by the index. Rewriting a page sitting at position 45 for AI Overview inclusion is work spent on the wrong constraint — get it into the top twenty first, then optimise the passage.

Assuming ChatGPT tactics transfer. They partly do and partly do not. The JavaScript rendering problem that dominates ChatGPT visibility does not apply here, because Googlebot renders. Conversely, the third-party corroboration that drives ChatGPT matters less. Applying one engine’s playbook wholesale to the other wastes half the effort.

Blocking snippets out of frustration. The nosnippet and max-snippet directives do restrict AI Overview use of your content — and they restrict ordinary snippets at the same time, which almost always costs more traffic than the AI Overview does. It is a lever that looks like control and behaves like self-harm for the large majority of sites.

If you lose an AI Overview citation

Citations rotate. A page cited last month and not this month has not necessarily done anything wrong, and the response depends on what actually changed.

First, confirm the overview still appears at all. Google suppresses AI Overviews on queries it classifies as sensitive or low-confidence, and it tunes those thresholds continuously. An entire category can lose its overview overnight, which looks identical to losing a citation.

Then check who holds the slot now. Whoever replaced you is showing you exactly what Google preferred — usually a more specific answer, a more current one, or a cleaner passage on the particular sub-question.

Then check whether your organic position moved. Because eligibility is gated by the index, a ranking slip frequently explains a lost citation without anything about the page changing.

One further check worth running: whether the query itself changed shape. Google periodically reinterprets what a query means, and a term that used to return a definitional overview may now return a comparative one. When that happens the sub-question you were cited for no longer exists, and the recovery is to write against the new sub-question rather than to improve the old passage.

A stored history makes this diagnosis take minutes rather than being unanswerable. Without a record of what the overview said last month and who it cited, you cannot distinguish between the feature changing, the competition changing, and your own position changing — and those need completely different responses. Our guide to diagnosing AI Overview disappearances covers the full checklist.

The checklist

1. Confirm the page ranks in the top twenty for its target query. If not, that is the work.
2. Check whether an AI Overview appears for the query, logged out and incognito. If not, this is not where your effort belongs.
3. Read the current AI Overview and note which sub-questions it answers and who is cited for each.
4. Find the sub-question you could own most cleanly and rewrite that section answer-first, in two or three self-contained sentences.
5. Add one specific, attributable fact to that passage — a dated number, a named source.
6. Add or correct structured data so it mirrors the visible text.
7. Re-check in three to six weeks, and read the trend rather than a single result.

Sources

• SERP feature tracking, 2026 — AI Overview presence averaging ~32.5% of monitored queries within a 31.02%–34.40% band.
• Pew Research Center — approximately 68,000 real search queries; clicks fell from 15% to 8% of visits when an AI summary was present.
• Ahrefs — CTR impact of AI Overviews; 34.5% decline for position one (April 2025), approximately 58% (February 2026).
• Google — 2.5 billion monthly active users for AI Overviews, stated 19 May 2026.

Frequently asked questions