All articles
AI Search Visibility

How to Rank in AI Overviews: What Actually Works in 2026

How to rank in AI Overviews in 2026: what Google actually rewards, what does not work, and how to check your AI impressions free in Search Console.

Armen Andonian Armen Andonian

In March 2026 Ahrefs looked at four million URLs cited inside Google AI Overviews and found that only 37.9 percent of them came from the top 10 search results. Eight months earlier, the same team put that number at 76.1 percent.

I lead with that because nearly every guide on this subject still opens with the same instruction: rank on page one first. That advice is not wrong. It is just no longer the whole job, and it skips the thing a UK business owner can genuinely act on this week.

So this is the version I give clients on how to rank in AI Overviews: what Google actually rewards, and how to check whether any of it is working without spending a penny. The things being sold to you that do nothing get their own section further down.

Can you finally see whether you appear in AI Overviews?

Yes, and if your business is in the UK there is a decent chance you got the tool before anyone else did.

On 3 June 2026 Google launched the Search Generative AI performance report inside Search Console, then widened access on 23 June. It gives you a dedicated view of impressions from AI Overviews, AI Mode and the generative AI features in Discover, broken down by page, country, device and date. The first wave went to a subset of website owners based in the UK, because the Competition and Markets Authority pushed Google on it.

There is one important limit. The report shows impressions only. No clicks, no click through rate, no query data. Google’s line when asked about clicks was that it is still working with website owners to understand which insights help most, with more metrics to come and no date attached.

Even so, this is the first honest scoreboard anyone has had for AI search, and it is free. Log in and look at it before you buy anything.

Google shipped a second control at the same time: a toggle that opts your site out of appearing in generative AI features. Google says using it does not affect your normal Search rankings, but an opted out site gets no traffic and no impressions from those features either. For almost every business I speak to, that is the wrong trade.

Do you still need to rank on page one?

It helps, and it is no longer enough on its own.

Here is the Ahrefs data in full. Of four million AI Overview URLs sampled in March 2026, 37.9 percent came from the top 10, 31.2 percent from positions 11 to 100, and 31.0 percent from pages that rank nowhere in the top 100 at all. Ahrefs put the shift down to Gemini 3, and to AI Overviews leaning less on the direct search results and more on the sources that surface in query fan out.

Query fan out is worth understanding, because it explains everything else in this post. When someone asks Google a question, the system does not run one search. It splits the question into sub questions, including ones the searcher never typed, runs them all at once, and assembles the answer from whatever those searches turn up. The page that ends up cited is often the one that answers the question next door.

A caveat before you rebuild your strategy on that 37.9 percent. The July 2025 and March 2026 studies used different sample sizes and different detection, so they are not a clean like for like comparison. The direction of travel is the reliable part, not the exact gap.

Where AI Overview citations come from now
Share of cited URLs by Google ranking position, March 2026
Top 10 results37.9%
Positions 11 to 10031.2%
Ranking nowhere in the top 10031.0%
76.1%
came from the top 10 in July 2025
37.9%
came from the top 10 in March 2026
Source: Ahrefs, 863,000 keyword SERPs and 4 million AI Overview URLs (March 2026); earlier figure from 1 million AI Overviews (July 2025). Different samples, so read the direction rather than the exact gap.

What actually gets a page pulled into an AI Overview?

Off site reputation, more than links. That is the uncomfortable finding.

Ahrefs studied 75,000 brands in August 2025 and measured how strongly different signals track with AI Overview visibility. Branded web mentions came out at 0.664. Backlinks came out at 0.218, roughly a third as strong. Brand anchors landed at 0.527 and brand search volume at 0.392. A follow up in December 2025 extended the work to AI Mode and put branded mentions higher again, at 0.709. Correlation is not causation and Ahrefs says so itself, but the top signals were all things happening away from your website.

On the page, the pattern is about substance you can check. Surfer analysed more than 405,000 searches and found that cited pages carry 38 percent more facts than pages that are not cited, and that pages with 10 or more facts get cited at twice the rate of pages with fewer than five. It is a vendor study rather than peer reviewed work, so weigh it accordingly. It also matches what I see: the pages that get quoted are the ones carrying named tools and real numbers with dates attached.

Formatting earns its keep too. Surfer found 78 percent of AI Overviews contain a list.

Two more findings worth knowing. YouTube is now the most cited domain in AI Overviews, and 18.2 percent of the citations that rank nowhere in Google’s top 100 are YouTube URLs. And freshness matters less than the panic suggests: SE Ranking found only 12.32 percent of cited sources were published in the previous 30 days, while 28.76 percent were published in 2025 and 26.85 percent in 2024. AI Overviews are not a news feed. A good evergreen page keeps earning.

How to rank in AI Overviews, step by step

Work in this order. It is sorted by what pays soonest for a small business, not by SEO textbook logic.

  1. Open Search Console and find the Search Generative AI performance report. Note your AI impressions by page and export them. That is your baseline, and without it every later judgement is a feeling.
  2. Write down the 20 questions your customers actually ask you, in their words, not yours. Longer questions matter most here: SE Ranking found queries of four words or more trigger an AI Overview 60.85 percent of the time.
  3. Search all 20 in an incognito window. Record which ones produce an AI Overview and who gets cited. It takes about half an hour and tells you more than most paid dashboards.
  4. Pick the five where a competitor is cited and you are not. Those are your targets.
  5. On each matching page, put a direct answer of roughly 40 words immediately under a heading phrased as the question itself. No preamble, no scene setting. The answer has to survive being lifted out of the page on its own.
  6. Add three checkable facts to that page, each with a date and a source, then add a list or a small comparison table. This is the fact density and formatting point from the Surfer data, applied.
  7. Re read the same Search Console report six weeks later and compare. Judge the work on impressions, not on whether you happened to spot yourself in one search.

Step two is where most people quietly cheat, writing the questions they wish customers asked. If you have an inbox or a phone log, use it. The real wording is already there.

What does not work, whatever you have been told?

Some of the most confident advice on this topic is flatly contradicted by Google’s own documentation.

Take llms.txt, the file some plugins and agencies now push. Google’s AI features guidance says you do not need to create machine readable files, AI text files, markup or Markdown to appear in Search including its generative capabilities, because Search does not use them. John Mueller went further and compared it to the old keywords meta tag, noting the Search team neither uses nor endorses it.

Special AI schema is the same story. Google states there is no special schema.org structured data you need to add for these features. Keep your FAQPage and Article markup for rich results in normal Search, but do not let anyone bill you for AI schema.

One genuine technical trap does exist, and it is the opposite of what people expect. If your pages carry a restrictive max-snippet value, or nosnippet, you can quietly remove yourself from AI features, since eligibility depends on the page being allowed to show a snippet in the first place. I have seen that setting left behind by an old plugin and never questioned.

Sold to you vs what Google actually says
The pitch What is actually true
"Add llms.txt so the AI can read your site"Google says Search does not use AI text files at all
"Special AI schema gets you cited"Google: no special schema.org data is needed for AI features
"We can get you placed in AI Overviews"No paid placement exists. There is nothing to buy
"Rank number one and you are in"Ahrefs found position one is a coin flip at best
"Search Console cannot track any of this"It can, since 3 June 2026, though impressions only

Will being cited actually bring you traffic?

Less than you would hope, and you should plan for that from the start.

Pew Research followed the browsing of 900 adults and found that when an AI summary appeared, people clicked a source link inside that summary in 1 percent of visits. They clicked any traditional result in 8 percent of visits, against 15 percent when no summary was present. Ahrefs measured the same squeeze from the other side across 300,000 keywords, finding the top ranking page saw a 34.5 percent lower click through rate when an AI Overview was on the page.

So treat AI Overviews as a brand visibility channel first. Being named as the source that answers the question changes who the buyer already trusts by the time they do click something, and that is the part that shows up later in enquiries rather than in analytics. It is the whole reason I built my search and AI visibility service around being cited rather than around ranking reports.

One more reason not to panic buy. Semrush tracked AI Overviews on 6.49 percent of queries in January 2025, 24.61 percent in July 2025, then 15.69 percent by November 2025 after Google recalibrated. The feature moves under your feet. Build pages that would be worth reading either way.

What does it cost to track AI Overview visibility?

Nothing, for the first six months, and I would keep it that way.

Search Console plus the incognito check from step three covers a small business properly. A spreadsheet with 20 questions, a date column and a yes or no is a perfectly respectable tracking system, and it beats a dashboard nobody opens.

When you do outgrow that, the entry point is Otterly.AI at 29 US dollars a month for 15 tracked prompts across AI Overviews, ChatGPT, Perplexity and Copilot. Semrush sells its AI Toolkit at 99 US dollars a month per domain on top of a base plan. Ahrefs Brand Radar is the serious option and priced like it, with the AI indexes charged per platform on top of a subscription, which realistically puts an all platform setup in the several hundred dollars a month bracket. Every one of those vendors prices in dollars or euros, so budget for the exchange rate and check how VAT applies before you commit.

My honest view, and it costs me work to say it: if you are a business with one site and a handful of service pages, buy none of them yet. Prices here were correct when I wrote this and this market changes fast, so check before you sign. There is more about how I work if you want the longer version.

What I would do on Monday morning

I would open Search Console, find the AI report, and just look. Most owners I speak to have never seen their AI impressions, and the first glance usually reframes the problem: either you are already being surfaced far more than you thought, or you are not in the conversation at all, and those two situations need completely different work.

Then I would take the five questions where a competitor gets cited and you do not, and fix those pages properly. A 40 word answer under a real question, a few facts with dates on them, a list where a list belongs. That is unglamorous work, and it is most of the job.

The wedge I keep coming back to is ownership. This is a skill you can hold yourself rather than rent forever, because the underlying habit, answering a real question plainly and backing it with something checkable, does not expire when Google ships its next model. If you want a hand working out which of your pages are closest to being cited, get in touch and I will tell you straight whether there is anything worth doing.

Frequently asked questions

Can I pay Google to appear in AI Overviews?

No. There is no paid placement, no submission form and no way to buy your way in. Google's own documentation says there are no additional requirements and no special optimisations needed to appear in AI Overviews or AI Mode. The only condition is that your page is indexed and eligible to be shown in Search with a snippet.

Do I need an llms.txt file on my website?

No. Google's AI features documentation says plainly that Search does not use AI text files, machine readable files or Markdown to decide what appears in its generative AI features. John Mueller compared llms.txt to the old keywords meta tag and said the Search team neither uses nor endorses it. If a guide is still telling you to add one, that guide is out of date.

Can I see how many clicks I get from AI Overviews?

Not yet. Since 3 June 2026 the Search Generative AI performance report in Search Console shows impressions from AI Overviews, AI Mode and AI features in Discover, split by page, country, device and date. There is no click data, no click through rate and no query data. Google has said more metrics will come over time but has not given a date.

Do I have to rank on page one to be cited in an AI Overview?

It helps, but it is no longer required. Ahrefs looked at four million AI Overview URLs in March 2026 and found 37.9 percent came from the top 10, 31.2 percent from positions 11 to 100, and 31.0 percent from pages ranking nowhere in the top 100. In July 2025 the top 10 share was 76.1 percent, so the pattern has moved a long way in eight months.

Why did my page disappear from an AI Overview it used to appear in?

An AI Overview is generated fresh for each search rather than stored as a fixed ranking, so the sources rotate. How often an AI Overview appears at all also moves sharply. Semrush tracked them on 6.49 percent of queries in January 2025, 24.61 percent in July 2025, then 15.69 percent by November 2025 after Google recalibrated the feature.

Does schema markup get me into AI Overviews?

Not directly. Google states there is no special schema.org structured data you need to add for its AI features. Schema is still worth having, because it earns rich results in normal Search and makes your pages easier for any machine to parse, but treat it as good housekeeping rather than a lever for AI Overviews specifically.

Should I block Google from using my content in AI features?

For most businesses, no. Search Console now has a toggle that opts you out of generative AI features, and Google confirmed that using it does not affect your normal Search rankings. The catch is that an opted out site receives no traffic and no impressions from those features at all, so you give up the visibility as well as the summarising.

How long does it take to get cited in an AI Overview?

Nobody has a confirmed figure, and anyone quoting you a precise number is guessing. AI Overviews are built from Google's index, so a new page has to be crawled and indexed and start earning some visibility before it can be pulled in at all. Plan in weeks rather than days, and judge it against your Search Console AI impressions rather than a gut feeling.

Armen Andonian

Written by

Armen Andonian

AI Automation & Search Visibility Consultant

I'm the founder of ACERO Digital, a London based SEO and digital PR agency. I help businesses cut manual work and scale with practical AI automation.

Keep reading