Original insights by Stephanie T. Wang, Jeffrey Gleason, Yakov Bart, Christo Wilson and Danaé Metaxa
30-second rundown
Key learnings:
- Google AI Mode cut external click-through by 18.8 percentage points: A randomized field experiment found that forcing people into AI Mode kept more attention inside Google without improving their experience.
- Removing AI features increased publisher visits: Participants shown Google without AI Overviews or AI Mode were 8.8 percentage points more likely to click an external result among those whose exposure actually changed.
- Visibility and traffic are separating: Businesses need pages that earn the smaller number of remaining clicks through original evidence, useful tools and details an AI summary cannot finish for the customer.
This week: Choose one important commercial topic and create something an AI summary cannot replace: original data, a calculator, an interactive comparison tool, a configurator or a useful template. Add it to the relevant page and make the title and opening clearly explain what visitors can do or discover there that they cannot get from the search result alone.
Coming 3 months: Build one zero-click-resistant page each month around a valuable customer question. Give every page a reason to visit, such as proprietary research, live data, an interactive tool or a genuinely useful resource, then measure qualified visits, tool usage, enquiries and assisted conversions rather than rankings alone.
Stephanie T. Wang, Jeffrey Gleason, Yakov Bart, Christo Wilson and Danaé Metaxa did something the search industry badly needed: they stopped guessing at the bloodstains and ran a randomized field experiment. Google AI Mode reduced external click-through by 18.8 percentage points. The doors stayed open, technically, but the casino moved the exits behind a velvet curtain.
The researchers changed the search experience for real users for seven days and found that Google’s conversational search kept attention inside Google while making users trust and enjoy it less.
1. The experiment put three versions of Google on trial
The researchers enrolled 1,444 US participants who primarily used Chrome and Google Search. After a three-day baseline, 1,100 people who searched during the experiment entered the behavioral analysis and 956 completed the final survey.
Participants were randomly assigned to three conditions. One group used ordinary Google with its current AI features. A second group had AI Overviews and AI Mode suppressed. The third was redirected into AI Mode for every Google search, like being handed a one-way ticket by a smiling clerk with no return address.
A custom browser extension recorded searches and outbound clicks. Random assignment provides stronger causal evidence than observing traffic trends and blaming the nearest robot.
Forced AI Mode reduced click-through by 18.8 percentage points compared with current Google Search. It also produced 2.27 fewer external clicks and 0.92 fewer search sessions per person per day. Removing AI features increased click-through by an estimated 8.8 points among participants whose AI exposure was successfully changed.
2. Google kept the visit and still made people unhappier
If AI Mode had reduced clicks by solving the problem brilliantly, Google could have argued that users traded wandering for efficiency. Instead, participants assigned to AI Mode reported lower trust, usefulness, satisfaction, agency and relevance.
They described difficulty reaching specific websites and seeing fewer, less diverse sources. Competitor-search use increased by 11.2 percentage points. Google built a room designed to keep people inside, then watched more of them look for another building.
The commercial lesson is that search visibility no longer guarantees a visit, even when the user dislikes the replacement. Ranking reports can look healthy while the traffic pipe narrows behind the wall.
A hotel page, for example, may still appear beneath an AI summary answering location, dining and room questions. The click now needs another reason: live availability, a room comparison, exact transfer details, original photography, a calculator or evidence the summary cannot safely compress.
3. Publishers must earn the click that remains
The study found fewer clicks to news sites, Reddit and Wikipedia under AI Mode. This is an economic warning for every company funding useful information with the expectation that discovery eventually produces a visit.
Stop treating rankings as the finished product. Separate search visibility, AI mentions, external clicks, assisted conversions and direct enquiries. A page can influence the answer without receiving the session, and a shrinking number of sessions may carry more intent than the old flood.
The response is not to bolt a desperate “click here” siren onto every paragraph. Publish things that remain valuable after the summary: proprietary data, interactive tools, detailed specifications, firsthand tests, downloadable templates, current inventory and clear commercial next steps.
The sample skewed younger and more educated than the US population, and forced AI Mode use may magnify reactions. But the direction is hard to ignore. Google is becoming the destination, not merely the road sign.
The blue-link bargain has been dragged behind the motel and robbed. Build pages worth leaving Google to visit, and stop celebrating visibility reports that never inspect the locked exit.
