68% of Google Searches End Without a Click: What the Data Really Says

For about 25 years, the deal with Google was simple. You published something useful, Google ranked it, and some of the people searching clicked through to your site. A growing number of searches now finish before that last step.

In June 2026, SparkToro published a study based on Similarweb clickstream data showing that 68.01% of US Google searches ended without a click between January and April 2026. Per 1,000 US searches, about 276 clicks reach the open web.

Those numbers have produced plenty of "SEO is dead" posts. The picture is more useful than that, and it's worth being careful about what the data can support.

What the numbers say

When a search does produce a click, it doesn't always leave Google's world. Search Engine Roundtable's summary of the chart splits those clicks like this:

Two other findings from coverage of the study matter:

• People search again. Searches that lead to another Google search rose 7.2 percentage points from 2024 to 2026. Part of the zero-click rise is people refining their question inside Google.
• AI Mode is still small here. Only 0.34% of searches moved into AI Mode during the period, so it doesn't explain the shift.

Three things this study doesn't prove

The jump may be overstated. The headline comparison is 68.01% against 60.45% in 2024, a rise of about 7.56 points. But SparkToro's studies have used different data providers over time. The 2024 number came from Datos and the 2026 number from Similarweb, and one report notes that historical comparisons should be made carefully. The direction looks real. The exact size of the change is less certain, so be wary of anyone calling it a precise two-year drop. The same applies to the widely quoted fall from 374 open-web clicks per 1,000 searches to 276.

It's US data. Coverage also says it excludes searches in the Google mobile app. We haven't found an equivalent recent study for India. Search behaviour here may differ, so use your own Search Console data before assuming your traffic will follow the same curve.

It doesn't prove a single cause. AI Overviews are a likely contributor, but the study doesn't isolate their effect. Rand Fishkin's own framing, quoted by Digital Applied, lists AI features, instant answers, interface elements and shifting user preferences together.

How much AI Overviews matter

The best-known measurement comes from Ahrefs. It compared 300,000 keywords between December 2023 and December 2025, and found that an AI Overview now correlates with a 58% lower average click-through rate for the top-ranking page. Its April 2025 version had found 34.5%.

The effect varies by position. Ahrefs puts the drop at about 58% in position one, 51% in position two and 19% in position ten. Keep in mind that this is correlation, that it's desktop data, and that it looks at informational searches. Many other searches, such as a brand name or a product someone is ready to buy, behave differently. AI Overviews also reportedly appear on more than 20% of Google searches, which leaves most searches without one, at least for now.

The practical lesson is that the damage is uneven. A page answering "what is customer churn" has a much bigger problem than a page comparing software for a specific job.

Is AI traffic really better?

You'll often see the claim that AI-referred visitors "convert 42% better". That comes from Adobe, and it needs context:

• It describes March 2026 data for US retail sites.
• Adobe's later reports show a 60% gap in July and a 90% gap in September, so the number moves quickly.
• A year earlier, the same visitors converted at around half the rate of other traffic.
• Adobe also sells a product for improving visibility in AI assistants, so it isn't a neutral party.

It's reasonable to think someone who read an AI answer and still clicked is further along in their decision. It's a mistake to assume that one retail number applies to a B2B software site or an Indian services business. Check your own conversion rate for AI referrals in analytics, even though the volume may be small.

What to change in your SEO

Sort your content by the kind of search it serves

Open Search Console and look at your top pages. Put each one in a rough group:

• Quick definitions and simple facts: at the highest risk of being answered on the results page. Don't stop answering these if they help your readers, but don't count on them for traffic.
• Comparisons, decisions and how-tos with real steps: the reader usually needs more than a summary, so these are often more durable.
• Brand searches: a measure of whether people remember you. No AI summary replaces someone looking for you by name.

Spend your next quarter of content effort mostly on the second group.

Publish something that can't be copied from page one

An article that rewrites the current top results gives nobody a reason to read it, or cite it. Original data, real examples, tests and hands-on experience are harder to replicate. You don't need a research department. Survey your customers, publish what your own numbers show or document a project honestly, including what went wrong.

Some research supports the broader idea. The academic paper that coined "Generative Engine Optimization" found that content changes could raise visibility in AI-generated answers by up to 40%, and that the results vary by subject. It was a controlled test on a benchmark of queries from 2023 and 2024, not a guarantee for your site. Be cautious with the many precise "citation lift" numbers you'll see quoted from it second-hand.

Make your pages easy to quote

This also helps human readers. Answer the question directly under the heading, then explain. Use tables for real comparisons. Name your sources. Keep titles and descriptions clear, because they still do the job of earning the click when one is available. Our free Google search preview shows how they could look.

Be findable beyond your own website

Buyers read reviews, forums, YouTube and LinkedIn, and those places can also feed AI answers. Useful participation in real communities and honest customer reviews help you whichever route people take. Be careful with claims that one forum or platform "trains" every AI model. We haven't seen reliable public evidence for most of the precise versions of that claim. And don't spam forums, since that tends to damage the reputation you're trying to build. Our piece on how search is splitting across social platforms and Google goes further into this.

Measure more than sessions

If you report only on organic sessions, you may panic over traffic that was never going to buy. Add these to your monthly view:

1. Branded search volume, from Search Console.
2. Impressions as well as clicks, because rising impressions with falling clicks tells you something different from a loss of rankings.
3. "How did you hear about us?" on your lead forms, with an option for AI assistants and a free-text box.
4. AI referrals in analytics, tracked as their own channel.
5. Pipeline or sales from organic, which matters more than any of the above.

You can also test your own visibility by asking a few AI assistants the questions your buyers ask, every month, and noting whether you appear. It's rough, and answers vary between runs, but it's cheap and it will show you who gets mentioned instead of you.

Search still pays, for fewer pages

A majority of US Google searches now end without a click, and pages that only answer simple questions are the most exposed. That's a real change in how search works. It doesn't mean organic search has stopped mattering: about 276 open-web clicks per 1,000 searches is still an enormous amount of traffic, spread across a smaller set of winners.

The sites that do well will give people a reason to click, and a reason to remember the name even when they don't. That was always good marketing. It just has less room for shortcuts now.

(Starting from the basics? Try our SEO audit checklist for small businesses or run your site through the free SEO audit tool.)

Want a search plan built around how buyers actually find you now? AgniCorp Media helps growth teams decide what to publish and how to measure it.