How to Rank in AI Overviews: What 24 Live SERPs Show About Who Gets Cited

There are two conversations about AI Overviews and they rarely meet. One is about lost clicks, backed by studies. The other is about “how to get in,” backed mostly by assertion. We wanted both in one place, so we took the click research at face value, then ran our own test to see which pages Google actually cites and on which queries.

This is the fourth article in our AI search cluster. If you want the wider map first, start with GEO vs AEO vs SEO. If you want the equivalent test for ChatGPT and Perplexity, it is here.

What Are AI Overviews Doing to Clicks?

Two independent measurements, one from outside the industry and one from inside it, point the same way.

Pew Research Center tracked 900 US adults through 68,879 real Google searches in March 2025. Around one in five searches produced an AI summary. When it did, users clicked a traditional result in 8% of visits, against 15% when it did not. Clicking a link inside the summary itself happened in 1% of visits. And people were more likely to simply stop: 26% of visits with a summary ended the browsing session, versus 16% without.

Ahrefs came at it from the ranking side. Comparing 300,000 keywords across March 2024 and March 2025, they found the presence of an AI Overview correlated with a 34.5% drop in click-through rate for the top-ranking page, from a forecast 4.0% to an observed 2.6%. Their sample was informational keywords only, because in their data 99.2% of keywords that trigger an overview are informational.

Google does not dispute that clicks move; it argues the trade is worth it. In May 2025 its head of Search wrote that AI Overviews were “driving over 10% increase in usage of Google for the types of queries that show AI Overviews.” Its documentation also states that traffic from AI Overviews and AI Mode is “included in the overall search traffic in Search Console” under the Web search type, which means you cannot separate it in your own reports.

So the honest baseline is this: on a query where an overview appears, expect roughly half the clicks you would have expected before, and expect not to be able to measure the change directly. Everything below is about being the page the overview cites, because that is the part you control.

What We Tested and How

On 10 September 2026 we ran 24 queries through Google Search, logged out, on desktop, with the interface set to English and the region set to the United States. Twelve were informational questions a marketing buyer might ask. Twelve were commercial, half of them generic (“ppc agency”) and half city-plus-service terms of the kind our own location pages target.

For each results page we recorded whether an AI Overview appeared, how many sources it showed, which sources they were, whether a local map pack appeared, and the top organic results. Then we checked whether the pages ranking in the top three were among the sources the overview cited.

Two limits. Our browser sat behind a Pakistani IP address even with the region set to the US, so the specific local businesses shown may differ from what a searcher in Houston sees. And no ads rendered in our sessions, so we cannot say anything about the ad-plus-overview layout. The patterns, not the individual domains, are the finding.

Which Queries Trigger AI Overviews?

Nearly all of them. Every one of the twelve informational questions produced an overview, from “what is local SEO” to “how much does PPC management cost.” That matches Pew’s finding that 60% of queries starting with a question word triggered a summary in March 2025, and suggests the rate has climbed since.

The surprise was on the commercial side. “PPC agency,” “PPC management services,” “best SEO agency,” “PPC management Orem,” “PPC management Baton Rouge” and “PPC management Houston” all showed an overview. These are queries an agency would call bottom-of-funnel, and Google put a two-thousand-character summary above the results anyway.

The exceptions had one thing in common. Of the six results pages that showed a local map pack, four showed no overview at all, and those four were exactly the city-plus-service queries: “SEO services Richmond VA,” “SEO company Houston,” “digital marketing agency Houston,” “local SEO company.” All eighteen pages without a map pack had an overview. When Google has a pack of verified local businesses to show, it appears to show that instead of a generated summary.

For a business with location pages, that reverses the usual worry. On your city terms the thing eating clicks above the organic results is not an AI Overview, it is the map pack, and the way into the map pack is a Google Business Profile, not content structure. We covered why that is the only local lever that works for us in our Houston page, and what it means for the other 57 cities in our location pages guide.

Who Gets Cited, and Does Ranking Still Matter?

On informational queries, ranking is most of the game. Across the twelve question SERPs, 20 of the 36 top-three organic results were cited by the overview above them. On “what is local SEO” and “how long does SEO take,” all three were. The overview is not pulling from a different universe; it is summarising the pages Google already trusts for that query, then adding a few more.

On commercial queries, ranking barely helps. Only 2 of 24 top-three results were cited, and both were the same agency on two Houston and Baton Rouge queries. Everywhere else the overview cited a definition or an explainer: Wikipedia and Salesforce for “PPC agency,” Indeed for “what is PPC management,” an agency’s pricing guide for “how much does PPC management cost.” The agencies ranking one, two and three were simply not in the summary. The reason is intent: a service page has nothing quotable to say about what PPC management is or costs, so the model quotes someone who does.

The overview cites people, not just publishers. Seven of the twenty overviews cited a Reddit thread, a LinkedIn post, a Quora answer or a named YouTube creator. On “how to get cited by ChatGPT,” two of the five citation groups were r/DigitalMarketing. First-hand experience written in a forum outranked, for citation purposes, most of the agency blogs on the same page.

A few domains are cited constantly. Search Engine Land appeared in 7 of 20 overviews. Indeed in 4. Semrush and LYFE Marketing in 3 each. Digital Rhodium in 0, which is consistent with our ChatGPT test and with a site that until this month had four articles and no answer-first pages on these questions.

How Do You Rank in AI Overviews?

Google’s own guidance is short: there are “no additional requirements to appear in AI Overviews or AI Mode,” and to be eligible a page “must be indexed and eligible to be shown in Google Search with a snippet.” Our data does not contradict that. It just tells you which pages, among the eligible ones, get picked.

1. Target question-shaped queries, and rank for them. Every informational query showed an overview and the overview drew most heavily on the top three. If you are not on page one for the question, none of the rest applies. This is classic SEO work and there is no shortcut around it.
2. Answer in the first paragraph. The cited pages opened with a definition or a direct answer. The overview is built from passages, and a passage that stands alone gets lifted. Our AEO guide covers the 40-to-55-word quick-answer pattern in detail.
3. Publish the explainer for your own commercial terms. On “PPC agency” the overview cited Wikipedia and Salesforce, not the agencies ranking. If you sell PPC management, the page that can be cited on that query is “What is PPC management and how does pricing work,” not your service page. Write it, link it to the service page, and let it carry the citation.
4. Stop optimising city-plus-service pages for the overview. On those SERPs it did not appear; the map pack did. Put that effort into the Google Business Profile and the signals that feed it.
5. Be present where the overview looks for first-hand experience. A third of overviews cited a forum, a LinkedIn post or a creator. A genuine, specific answer on Reddit or LinkedIn from a named person at your company is a citation path most agencies ignore.
6. Keep structured data honest, and do not expect it to do this job. Google asks that “structured data matches the visible text on the page.” Nothing in our sample suggested schema drove a citation. Our schema guide explains what it does instead.
7. Decide, page by page, whether you want to be in it. Google honours nosnippet and max-snippet for AI features. If a page’s value is the click, not the mention, you can opt that page out without touching the rest of the site.

How Do You Measure the Effect on Your Own Traffic?

You cannot isolate it in Search Console, because Google folds AI feature traffic into the Web search type. What you can do:

1. Watch click-through rate by position for question queries. In Search Console, filter to queries starting with what, how, why, which. If CTR at positions one to three falls while impressions hold, an overview is probably absorbing the click. Compare against your commercial queries, which will show a smaller effect.
2. Keep a query panel. The same fixed list we used for the ChatGPT test works here. Once a month, note whether an overview appears and whether you are cited. Presence changes; ours will not look identical next month.
3. Count the map pack separately. For local terms, record whether a pack appeared and whether you are in it. That, not the overview, is the number that predicts local leads.
4. Report mentions and clicks together. A citation without a click still puts your name in front of the buyer. Pew found 1% of visits click into the summary’s links, so the value is mostly the mention. Treat it like a PR placement, not a landing page.

Frequently Asked Questions

← Back to all articles