AI Overviews are taking the click. The numbers, and what UK retailers should do about them.

For two years "AI is killing our traffic" has been a feeling. Now there is a measurement behind it. A controlled field study found that when Google showed an AI Overview, the share of searches that ended without a single click rose sharply, and the clicks that did leave Google fell by nearly half. The sample was American, so the exact numbers need a UK pinch of salt, but the shape of the change is hard to ignore. The question this article answers is not how to claw the clicks back. It is how to stop measuring the wrong thing.

What the study actually found

Most claims about AI Overviews are guesswork stitched together from traffic dips. This one is not. Researchers at the Indian School of Business and Carnegie Mellon University ran a controlled experiment with 1,065 US participants over January and February 2026, with each person tracked for two weeks. Because it is a randomised setup rather than a correlation drawn from analytics, it gets closer to cause and effect than the usual dashboard-watching.

Three figures matter.

• AI Overviews appeared on 42% of the queries in the study. This is not a fringe feature you can wait out. It is already on nearly half of searches.
• When Overviews were removed, outbound clicks rose from 0.38 to 0.61 per search. Turned the other way round: when the AI summary is present, the clicks leaving Google to visit a website drop by roughly 38%.
• Zero-click searches fell from 72% to 54% when Overviews were removed. So with an AI Overview on the page, nearly three quarters of searches ended without the person clicking through to anything at all.

It is a US sample, so treat the precise percentages as directional for the UK rather than gospel. But the direction is not subtle, and there is no reason to expect the UK to be immune. Fewer people are clicking through, and the AI summary is the reason.

Why chasing the lost clicks is the wrong move

The instinctive response is to try to win the clicks back: more content, more keywords, more SEO effort aimed at the same shrinking pool of outbound visits. That is running harder in a direction the platform is deliberately closing off.

The clicks are not coming back to their old level, because the AI Overview is doing exactly what Google designed it to do, which is answer the question on the page. Pouring budget into recovering a metric that the platform is actively suppressing is how you spend a lot and feel like you are losing ground.

The better response is to change what you count. If a smaller number of people are arriving on your site, the question that matters is not "how do we get the old volume back", it is "how much profit does each visit that does arrive actually produce, and is that going up or down". A retailer measuring sessions will read this shift as a disaster. A retailer measuring profit per visit will see what is really happening, and can act on it.

“The kneejerk reaction is to win the clicks back with more content and more SEO aimed at the same shrinking pool. That’s the wrong move. A click was always a proxy for a profitable customer — so the job now is to get visible inside the AI answer and measure what each visit is actually worth, not to chase a metric Google is deliberately suppressing.”

— Ross Miles, Coffee Marketing

The two things worth doing instead

1. Be the source the AI Overview quotes

When an AI Overview answers a shopping question, it cites sources, and being one of those cited sources is the new front row. It keeps you visible in the exact place the click used to come from, and it lends you authority even when the person does not click, because your brand is named inside Google's own answer.

Getting cited is not luck. It comes from the same foundations that AI-led search rewards everywhere: clean, complete product data, structured data (schema) on your pages that makes them machine-readable, genuinely useful content that answers the questions people ask, and the technical files that let AI tools crawl and quote you. This is where the SEO effort should go now, towards being quotable by the AI, not towards fighting it for a click it is designed to keep. We unpack how AI decides who to show in our guide to the 2026 search shift.

2. Measure profit per visit, not sessions

If fewer, better-qualified people are landing on your site, your reporting has to be able to see that. Most ecommerce measurement cannot, because it treats every session and every conversion as equal. A £10 order and a £100 order signal the same thing to Google, refunds and returns rarely flow back into the numbers, and traffic volume gets treated as the headline when it is really a vanity metric.

Fix the measurement and the AI-Overview story changes from panic to clarity. Track profit-based conversion values, feed returns and refunds back in, and report profit per visit rather than raw sessions. Do that, and a fall in traffic that comes with a rise in profit per visit is revealed for what it is, which is often fine, or even good. Leave the measurement as it is, and you will make defensive decisions based on a number that was never telling you the truth in the first place.

The uncomfortable but freeing conclusion

The click was always a proxy. What you actually want is profitable customers, and the click was just the route most of them used to take. As that route narrows, the businesses that struggle will be the ones who kept optimising for the proxy. The businesses that do well will be the ones who got visible inside the AI answer and measured what the visit was worth once it arrived.

You cannot control how many clicks Google chooses to release. You can control whether you are the retailer it quotes, and whether your numbers tell you the truth about the visits you get. Those are the two levers worth pulling.

What we do, and where to start

We help UK retailers stay visible as search shifts to AI, and measure what actually matters once the visit lands.

A good first step is an AI Visibility Report, which tests where you show up when a shopper asks an AI a question in your category, whether you are being cited, and who is getting recommended instead. If the deeper issue is that your reporting cannot tell a good visit from a bad one, a Tracking Audit sorts out profit-based conversion values, returns and refunds, so you can measure profit per visit rather than mourning lost sessions.

The traffic shift is real. The right response is not to fight it for clicks, but to get quoted and to measure properly. If you want a second opinion on where you stand, tell us the challenge and we will take a look.

Frequently asked questions

Are these AI Overviews figures from the UK?

No. The study tracked 1,065 US participants in early 2026, so treat the exact percentages as directional for the UK. The pattern, fewer outbound clicks and more zero-click searches when an AI Overview shows, is what to take from it.

Does this mean SEO is dead?

No, it means the target has moved. Instead of optimising to win a click, you optimise to be the source the AI Overview quotes, using clean product data, schema and genuinely useful content. The work is similar, the goalpost is different.

How do I know if my traffic drop is actually a problem?

By measuring profit per visit, not sessions. A fall in visits alongside steady or rising profit per visit is usually fine. You can only see that if your tracking values conversions by profit and accounts for returns, which most ecommerce setups do not do by default.