Business SystemsPerspective
Google's AI Mode Cut Clicks to Websites by 18.8 Points in a Controlled Test. What a Small Business Site Should Do Now.
For two years the argument about AI search and website traffic has been fought with before-and-after charts and company statements. In August a university team published a controlled experiment. The number is large, the caveats matter, and the useful response for a small business is less about SEO tricks than about being the source an answer has to rely on.
Author
DueClix Engineering
Published
Reading time
6 min read
Whether AI answers in Google take visitors away from websites has mostly been argued with correlations and with Google's own statements. In August 2025 Google's head of Search wrote that "total organic click volume from Google Search to websites has been relatively stable year-over-year," and that the clicks being sent were of higher quality.
On 18 August 2026 researchers from the University of Pennsylvania and Northeastern University posted something of a different kind: a preregistered field experiment titled "AI in Search Reduces Publisher Referrals Without Improving User Experience." The search-industry press picked it up in the second half of September. It is the first result we have seen that measures cause rather than correlation, and every small business with a website should know what it found and what it did not.
What the experiment did
1,100 people in the United States, recruited between 17 and 19 March 2026, installed a browser extension that recorded their Google searches. After three days of searching as normal, each was assigned for seven days to one of three versions of Google: unchanged, where AI Overviews may appear and AI Mode is one tab away; no AI, with AI Overviews hidden and AI Mode removed; or AI Mode only, where every search was redirected into AI Mode.
The outcome that matters to a site owner is click-through rate, which the paper defines as external clicks from Google divided by Google searches. In other words: how often a search ends with someone visiting a website that is not Google.
| What was measured | Effect of AI Mode |
|---|---|
| Share of searches that ended in a click to another website | 18.8 percentage points lower |
| Google search sessions per day | 0.92 fewer |
| Share of users who clicked through to news sites | 12.5 points lower |
| Share of users who clicked through to Reddit | 21.2 points lower |
| Share of users who searched on Bing, DuckDuckGo or Yahoo | 11.2 points higher |
| Trust in information found on Google (7-point scale) | 0.34 points lower |
The other direction is just as telling. Removing AI Overviews and AI Mode raised click-through by 8.8 points compared with unchanged Google. So the default Google that most people use today, with AI answers above some results, already sends fewer visitors than a Google without them. Participants pushed into AI Mode also rated search as less useful and less satisfying, which undercuts the idea that fewer clicks simply means people got what they needed.
What it does not show
The authors are careful about the limits, and so should anyone quoting the headline number be. Before the experiment, participants used AI Mode for 0.6 percent of their searches, so forcing it onto every search describes full adoption, not today's behaviour. The sample skewed younger, more educated and more left-leaning than the US population. Compliance in the no-AI arm faded to roughly half over the week, and seven days says nothing about how habits settle over months.
Most importantly, 18.8 is a change in the share of searches that ended in a click. It is not a forecast that any particular website will lose 18.8 percent of its visitors. What the study does establish is direction, with a cause attached: when the answer is written on the results page, fewer people leave it.
Owners are already noticing
A Small Business Expo survey of 3,608 small business owners, published on 29 September, found that 33.9 percent had noticed a decline in website traffic they believe is related to AI-generated answers in search, and 12.9 percent described that decline as significant. 42.8 percent had noticed no decline, and 4.1 percent said their traffic had gone up.
The figure that should worry an owner more is a different one: 14.8 percent said Google or another AI-powered search result had shown incorrect information about their business. This is self-reported perception, not measured traffic, but it points at the part of the problem a small business can actually do something about.
What a small business website should do now
Start with what Google itself says. Its documentation for site owners states that "there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary," and that a page must be indexed and eligible to be shown with a snippet to be used as a supporting link. That sentence is worth remembering the next time someone offers an "AI SEO" package. The work that remains is ordinary, and it matters more than it did.
- Make sure you can be cited at all. A page that is not indexed cannot be a supporting link. In Search Console, check that your service, location and contact pages are actually indexed, not sitting in "Discovered – currently not indexed". Some of our own newest pages sat in that state for weeks in September; it is common, and it is the first thing to check.
- Write down the facts an answer needs, as plain text on your own site. Hours, service area, what you do and what you do not do, how pricing works. An answer cannot repeat a correct fact that is only in a photo of a flyer or nowhere at all.
- Make every page say the same thing everywhere. Title, meta description, structured data and the visible page should agree. We have written about treating a site's SEO as a data model so that they cannot drift apart.
- Ask the AI about yourself once a month. Search your business name and your main service in AI Mode and in the AI assistants your customers use. Note anything wrong, then correct the sources it is likely to draw on: your own pages, your business profile and the directories that list you.
- Measure leads, not only visits. Google counts AI feature traffic inside the ordinary Web totals in Search Console, so it cannot be separated out there. Track enquiries from your contact page as a conversion, so a fall in visits can be judged against what actually reaches you.
Be the page an answer has to rely on
An AI answer can summarise a generic explainer well enough that nobody clicks. It cannot know your prices, your service area or whether you can take a job next week unless you have written them down, and when it guesses, it can be wrong, which about one owner in seven in that survey says has already happened to them. The pages that keep earning the visit are the ones holding information only you have: specific, first-hand and current.
A controlled experiment now says AI answers take clicks away. The response for a small business is not to chase the answer box. It is to make sure that whatever the answer says about you is correct, and that the page it is correct because of is yours.
Sources
- Stephanie T. Wang, Jeffrey Gleason, Yakov Bart, Christo Wilson and Danaé Metaxa, AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence, arXiv 2608.18352, submitted 18 August 2026. Design, recruitment dates, every figure in the table, the 8.8-point no-AI result, the 0.6 percent baseline and the stated limitations.
- Search Engine Journal, Research Shows Google AI Mode Sends Less Clicks & Is A Poor User Experience, 21 September 2026.
- TechWyse, Study: Google AI Mode Cuts Clicks, Drives Users to Rivals, 22 September 2026.
- Small Business Expo Research Team, AI Search: 1 In 3 Owners Report Falling Website Traffic, 29 September 2026. Survey of 3,608 small business owners.
- Google, Liz Reid, AI in Search is driving more queries and higher quality clicks, 6 August 2025.
- Google Search Central, AI features and your website, last updated 10 December 2025. Eligibility, the no-special-optimisation statement and Search Console reporting.