AI Overviews Are Reducing Clicks: What Website Owners Should Optimize

You've noticed that your rankings have not moved but your traffic has. That gap is what sent most people looking for an explanation, and what they find is a pile of scary percentages that all seem to say something slightly different.

This article does three things. It sorts out what the numbers actually measure, it walks through a quick check to confirm AI Overviews are really your problem, and it lists what is worth your time now, straight from Google's own guidance. No tools to buy.

 

The three important AI Overviews stats and findings

Three numbers get thrown around as if they were the same finding. They are not. Here is what each one actually measures, so you don't spend a weekend fixing the wrong thing.

 

What the 58% measures

It is a click-through rate, and only for the top result. Ahrefs re-ran its study on December 2025 data and published the update in February 2026. It found that when an AI Overview appears, the top-ranking page gets about 58% fewer clicks.

 

What the 68% measures

This one is about searches, not about your site. SparkToro looked at Similarweb clickstream data from January to April 2026 and found that 68.01% of US Google searches ended without a click.

SparkToro has tracked zero-click search behavior for years, and here's what they found:

• 2019: 50,33% of Google searches ended without a click
• 2020: 64.82% of Google searches ended without a click between January and December.
• 2024: 58.5% of Google searches in the U.S. (and 59.7% in the EU) ended with no clicks.

 

What the 38% measures

This one is about citations in AI Overviews. The 38% means that only 38% of pages cited in an AI Overview also rank in Google's top 10 for that query. The rest rank below the first page — or don't rank in the top 100 at all.

The takeaway is simple: ranking in the top 10 is no longer the only path to being cited. Covering a topic thoroughly and addressing related questions can give your content more opportunities to appear.

Figure 1 puts all three numbers side by side.

Overview: Three very different numbers — 58%, 68%, and 38% — measure different things. One measures the drop in clicks to the top result when an AI Overview appears, another measures how often Google searches end without a click, and the third measures how often AI Overview citations come from pages ranking in the top 10.

 

Why ranking first stopped guaranteeing an AI Overview citation

 

How query fan-out works

Google confirms that its AI features use query fan-out. The model takes your search and quietly runs a set of related ones alongside it. Google's own example: "how to fix a lawn that's full of weeds" fans out into searches like "best herbicides for lawns".

So your page can get cited for a question you never tracked.

 

What that did to citations

Ahrefs looked at 863,000 keywords and 4 million AI Overview URLs. Only 38% of cited pages also appeared in the top 10 for that same query, down from 76% in July. The rest split almost evenly: 31.2% ranked somewhere between 11 and 100, and 31.0% did not rank at all.

Here is the practical read. Cover a topic across related angles instead of guarding one ranking. The same thinking drives getting found in ChatGPT and other AI search platforms. Figure 2 shows how far things moved.

 

Before you optimize, confirm it is AI Overviews

Rebuilding a site over a seasonal dip is an expensive mistake. Four checks, using Search Console and Analytics only, will tell you whether AI Overviews are the likely cause. Figure 3 walks through them in order.

• Impressions versus clicks. If impressions hold steady while clicks fall, people are still seeing you and something is eating the click. That fits AI Overviews. If both fall together, you have a ranking or indexing problem.
• Query mix. AI Overviews show up mostly on informational searches. If your branded and buying queries are flat and only your how-to pages dropped, that fits too.
• Core update timing. Check the date of your drop against Google's published list of ranking updates before you blame AI.
• Your own changes. Redesigns, migrations, and one stray robots.txt edit explain more traffic drops than AI does. Google's guide to debugging search traffic drops covers the same ground.

 

The five things worth optimizing now

All five come straight from Google's own guidance. They are ordered by effort against payoff for a small site.

• Stay eligible. To show up in AI features, your page has to be indexed and allowed to show a snippet, and your site has to be included in Search generative AI features in Search Console. Check that no stray nosnippet or max-snippet tag is quietly hiding you.
• Fix crawling first. Google asks that crawling is allowed in robots.txt and by your CDN or host, and that your important content is there as text. If you have not looked in a year, run a technical SEO audit before you touch anything else.
• Write something only you could write. Google draws the line with its own example: "7 Tips for First-Time Homebuyers" is common knowledge, while "Why We Waived the Inspection & Saved Money" brings real experience. One first-hand page beats five summaries.
• Add real images and video. AI features can pull in relevant images and video, which gives your site more ways to appear than web links alone. It matters more than it sounds: among citations that did not rank in the top 100, 18.2% were YouTube URLs.
• Keep your business details current. AI answers can include product listings and local business information, so your Merchant Center feed and Google Business Profile help you show up. It is a 20-minute job most small sites never revisit.

 

What Google says you can ignore

This section reflects Google's own published guidance on generative AI search. You may come across AEO advice elsewhere that appears to contradict it, especially around tactics such as llms.txt, content structure, and AI-specific optimization. Where there is conflicting advice, we've focused here on what Google itself says it uses or recommends.

Google's mythbusting section on generative AI search names five popular tactics as unnecessary. Figure 4 puts them next to the five actions above.

• llms.txt and special markup. You don't need new machine readable files, AI text files, or markup, because Google Search does not use them. Keeping one for other tools is fine and changes nothing on Google.
• Chunking content. There is no need to break pages into tiny pieces, and no ideal page length.
• Rewriting just for AI. AI systems understand synonyms and general meaning, so you don't have to capture every phrasing someone might use.
• Buying mentions. Chasing inauthentic mentions is not as useful as it sounds, since ranking systems favour good content and other systems block spam.
• Schema as an AI requirement. Structured data is not required for AI search and there is no special markup to add. Keep using it for rich results.

Our take: If an agency is charging you to create an llms.txt file or rewrite existing pages specifically for AI, ask what Google guidance that work is actually based on. Google says these tactics are not required for AI search. Your budget is likely better spent creating one genuinely useful, first-hand article that adds information, experience, or insights your competitors don't have.

 

Measure visibility, not just sessions

 

What Search Console shows you now

Clicks alone no longer tell you whether you are winning. Google now gives you a free tool for the other half: the Generative AI performance report in Search Console, which shows how your content is doing in AI features on Search and Discover.

 

What it still does not show

Sites appearing in AI features are already counted in the main Performance report under the "Web" search type, so you cannot pull AI Overview clicks out on their own.

Google also says clicks from pages with an AI Overview are higher quality, with people more likely to stay a while. That is Google's own claim and no independent study backs it, so check it against your own conversion data. Swap your headline number from sessions to impressions plus conversions, and the picture you get is much more clear.

 

A realistic week for a small team

Monday, run the four checks and confirm what you are actually dealing with. Tuesday, check crawling and snippet eligibility, then fix whatever is blocking you. Wednesday and Thursday, write one page on something you genuinely know, with your own photos. Friday, open the Generative AI performance report and save a baseline. That is the whole plan.

If domain admin is what eats your Friday, Dynadot MCP lets you manage domains by talking to an AI assistant instead of clicking through DNS panels.

 

Frequently asked questions

 

Do I need an llms.txt file to appear in AI Overviews?

No. Google states that you don't need new machine-readable files, AI text files, or markup to appear in Google Search, including its generative AI features, because Search doesn't use them. Keeping one for other tools is fine and has no effect either way on Google.

 

Does schema markup help me get cited in AI Overviews?

Structured data isn't required for generative AI search and there's no special schema type to add. It's still worth maintaining for rich results in regular Search. If your markup doesn't match the visible text on the page, fix that first.

 

Can I see how many clicks AI Overviews actually sent me?

Not separately. AI features are folded into the overall Search Console Performance report under the "Web" search type. The Generative AI performance report shows how your content performs in those features, so use it for visibility and use Analytics for conversion quality.