Skip to content
Marketing, IT & AI automation · Surrey and Metro Vancouver
Digital Allo Marketing Services Ltd. Contact Us

Digital Allo Blog · November 19, 2026

Google’s AI Answers Now Cost Your Website 40% of Its Clicks — What Surrey Businesses Do Instead

By Digital Allo Editorial Team Published

Your website can rank number one on Google and still get far fewer visitors than it did a year ago. Until now, that was a suspicion backed by correlations. This week it became a measured fact: a randomized experiment found that when Google shows an AI Overview, clicks to outside websites fall by about 40%.

For a Surrey business that wins customers from Google, that changes what ‘ranking well’ is worth — and what to do about it.

The first causal number on the missing clicks

Researchers Saharsh Agarwal (Indian School of Business) and Ananya Sen (Carnegie Mellon) ran a pre-registered field experiment on 1,065 US desktop Chrome users in January and February 2026. A browser extension randomly hid AI Overviews for some of them, then watched two weeks of real searches. On searches that would show an Overview, the people who saw it averaged 0.37 clicks to outside websites per search. The people who did not see it averaged 0.62. The authors report the difference as a 39.8% reduction.

Sessions that ended with no click to any outside website rose from 54% to 73% when the Overview was visible. Across all searches — including the ones that never trigger an Overview — the researchers estimate the feature cut organic Google clicks to websites by about 18.5%.

Two details matter as much as the headline. First, ad clicks did not change: the lost clicks were organic, not paid. Second, the people who saw the AI answers were no more satisfied. A closing survey found no measurable difference in satisfaction, perceived quality, or ease of finding information — and more than 95% of the treated users never noticed the Overviews had been hidden from the others. The searcher loses nothing. The website loses the visit.

A second, separate preregistered experiment — Wang and colleagues at Penn and Northeastern, 1,100 participants — reached the same conclusion from the other direction: removing AI Overviews and AI Mode increased clicks to publishers. Earlier observational work from Pew Research Center (July 2025) found users clicked a conventional result on 8% of visits to pages containing an AI summary, against 15% on pages without one — but that comparison could not separate the summary from the kind of search. The randomized design is what makes this one different: it is the first result that answers Google’s standard reply, which has been that correlation is not causation.

How much of search now carries an AI answer

The effect is growing because the format keeps spreading. DataForSEO’s panel of 20 million tracked US keywords showed AI Overviews on 61.43% of keywords in August 2026, up from 5.93% in September 2024. Coverage is highest for health and finance keywords (roughly 70%) and lower for dining and real estate. Featured snippets — the old answer box — have nearly disappeared in the same period.

Read those coverage figures as vendor measurements, not Google’s own numbers, and read the experiment with its limits in mind: it is a working paper that has not yet completed peer review, it covers US desktop Chrome users only, it ran for two weeks, and its statistically significant result applies to informational searches — the samples for navigational and transactional searches were too small for firm conclusions. The direction of the finding, though, now has two randomized experiments behind it, and the 18.5% figure is an estimate for US desktop, not a global number.

What it means for a Surrey business

Informational searches are the ones that bring new customers in: ‘how much does a new roof cost in Surrey’, ‘best accountant for a small business’, ‘what should a website cost’. Those are exactly the searches where an AI Overview now answers the question on the results page — and where, in the experiment, nearly three in four sessions ended without a click to any website.

Three consequences follow:

Ranking first is worth less than it was. The position still matters — it decides which pages the AI can cite — but the traffic that position used to deliver is no longer guaranteed. A business that judges its SEO by website visits alone will see the numbers fall and may wrongly conclude its rankings fell too. The November 2026 core update is still rolling out (started 10 November, expected to finish around 24 November), so do not judge any ranking impact until at least a week after it finishes — and do not confuse a traffic dip caused by AI answers with a ranking drop caused by the update.

The citation is the new click. When the answer names your business, links to your page, or quotes your price list, you are in the conversation even if nobody clicks through. When it does not, you are not. That is the same conclusion our earlier posts reached from different evidence: Blog 03 showed reviews decide which businesses AI recommends, and Blog 13 showed 74% of businesses are never named at all. This experiment adds the cost side: being unnamed now costs you a visit you used to get for free by ranking.

The action matters more than the visit. The AI answer can carry your phone number, your booking link, your hours. A customer who reads the answer and calls you directly never appears in your website analytics — and never needed to.

What to do this week

  • Write pages the AI can quote, not just pages that rank. Open each key service page with a one-sentence direct answer, then prove it with real numbers from named sources. Pages built to be cited are the ones that get cited — and a citation in the answer is now worth more than a blue link below it.
  • Win the action inside the answer. Make sure your Google Business Profile has a working booking link, a click-to-call number and current hours, and that the same facts match on your website. If the answer can send the customer straight to a booking or a call, the missing website click stops mattering.
  • Build one channel you own. An email list, a repeat-customer list, a profile your customers follow — traffic that does not pass through a Google results page at all. Every customer who comes back directly is one the AI summary cannot take away.
  • Measure citations, not just clicks. Search Console’s generative AI performance report shows impressions in AI Overviews and AI Mode by page. Once a month, ask ChatGPT, Google AI Mode and Perplexity the exact questions your buyers ask, and note which firms and sources come up. Track whether you are being named — that is the metric the click numbers no longer measure.
  • Don’t buy a ‘GEO quick fix’ for this. The study’s own conclusion is that there is no separate remedy: being discoverable in AI answers still largely depends on familiar search signals, presence on third-party sources and pages written clearly enough to be cited. Spend the money on reviews, on pages worth citing and on a profile worth booking from — the same work Blog 03 recommended, now with a measured reason to do it.

Google’s AI answers did not make search useless. They made the click optional — and the businesses that adapt are the ones that get named in the answer, get the call or the booking from it, and keep a direct line to their customers for the next time. The ones that only chase rankings will keep winning positions and wondering where the visitors went.

Call Digital Allo at (604) 401-0699 or visit https://www.digitalallo.com/contact-us/ — 13898 64 Ave #108, Surrey, BC.

A note on the numbers in this post: the 39.8% figure, the 0.37 vs 0.62 clicks, the 54% to 73% zero-click sessions and the 18.5% estimate all come from Agarwal & Sen’s working paper as reported on 18–19 November 2026; the coverage figures come from DataForSEO’s keyword panel; the Pew figure is observational. The experiment is a working paper, not yet peer reviewed, covering US desktop Chrome users over two weeks in early 2026 — read the numbers as the best causal evidence available, not as exact figures for Canadian searches.

FAQ

How much traffic do AI Overviews actually cost websites? In the Agarwal & Sen randomized experiment, searches that showed an AI Overview produced about 40% fewer clicks to outside websites (0.37 clicks per search against 0.62 when the Overview was hidden), and sessions ending with no click at all rose from 54% to 73%. Across all searches, including ones that never trigger an Overview, the researchers estimate about 18.5% fewer organic clicks.

Is this proven, or just another correlation? This is the first causal result on the question: a pre-registered randomized field experiment on 1,065 US desktop Chrome users, where a browser extension randomly hid AI Overviews. A second preregistered experiment (1,100 participants) found the same direction. The limits: a working paper not yet peer reviewed, US desktop only, two weeks of searches, and firm conclusions only for informational searches.

Do AI Overviews hurt ads as well as organic clicks? No — the experiment found no corresponding change in advertising clicks. The lost clicks were organic. That is one reason Google can keep expanding the format while the cost falls on websites, not on its ad revenue.

Does this mean SEO is dead? No. The study’s own conclusion is that being discoverable in AI answers still largely depends on familiar search signals, presence on third-party sources and pages written clearly enough to be cited. Ranking still decides what the AI can quote — but the traffic a ranking delivers is no longer what it was, so citations and direct actions matter more.

What should a Surrey business do first? Check whether your business is being named in AI answers for the questions your buyers actually ask, make sure your Google Business Profile can take a booking or a call directly, and write service pages that open with a direct answer the AI can quote. Then track citations monthly — not just website visits.

Service areas

Serving businesses across British Columbia

On site and remote across Metro Vancouver and the Fraser Valley. Remote-first on Vancouver Island and in the Okanagan, with visits by arrangement.

Surrey & South of the Fraser

On-site and remote

Vancouver, Richmond & North Shore

On-site and remote

Burnaby, Tri-Cities & Maple Ridge

On-site and remote

Fraser Valley

On-site and remote, with planned visits further east

Vancouver Island

Remote-first, visits by arrangement

Okanagan & Thompson

Remote-first, visits by arrangement

See all service areas