Something has shifted in the way people find businesses online, and most owners haven't been told. For two decades the deal was simple: rank well on Google, earn the click, win the visitor. That arrangement is quietly coming apart. By early 2026, according to SparkToro's analysis of Similarweb clickstream data, 68% of Google searches in the US ended without a single click - up from around 60% in 2024. The search still happens; the visit to your website increasingly does not. For UK small businesses that have spent years and budgets chasing page-one rankings, this is either an existential worry or the best opportunity in a decade. We think it's the latter, provided you understand what has actually changed.
The traffic that quietly disappeared
When Google places an AI Overview at the top of the results - its machine-generated summary that answers the query on the page itself - click-through rates to the links below drop by close to 60%, per the same SparkToro research. AI Overviews now appear on more than one in five searches. Google's newer AI Mode, a full conversational search experience, passed a billion monthly users in 2026 with query volume more than doubling each quarter. The practical effect is that a large share of the informational questions your customers used to type into Google - "how much does a new website cost", "do I need an accountant for a limited company" - are now answered before anyone reaches a result. The click you were counting on never arrives. This is not a penalty or a glitch; it is the new default, and no amount of traditional keyword tuning reverses it.
Why the clicks you still get are worth more
Here is the part the doom-laden headlines miss. The visitors who do arrive are markedly better. Semrush found that traffic referred from AI search converts at roughly 4.4 times the rate of traditional organic search, and Ahrefs reported a case where just 0.5% of visitors from AI drove 12.1% of sign-ups. SE Ranking's study of more than 100,000 websites found AI-referred visitors spend 67.7% longer on a site - an average session of nine minutes and nineteen seconds against roughly five and a half minutes for organic. The reason is intuitive: an AI assistant has already done the sifting, so by the time someone clicks through to you they are further down the decision path and closer to buying.
The pattern holds closer to home, too. SE Ranking's figures put UK AI-referral traffic up 21% year on year, at around 0.16% of total traffic - smaller than the US, but climbing steadily, and the businesses that get cited now build an advantage before their competitors have noticed the shift. Raw traffic is falling, but the quality of what remains is rising sharply. The winners will stop measuring success by volume of visits and start measuring it by whether they are the name the AI recommends.
How to earn a place in the answer
Getting recommended by an AI is a different craft from ranking on Google, though the two overlap. The emerging discipline goes by a few names - answer engine optimisation, generative engine optimisation - but the fundamentals are consistent. Structured data matters more than ever: clear schema markup that tells machines exactly what your business is, what it sells, and who it serves. Google's knowledge panels, which sit largely untouched by AI Overviews, are built from precisely this kind of entity markup. Content should be written to answer, not to pad - a page that opens with a direct, quotable answer to a real question is far more likely to be lifted into an AI response, and SparkToro's data shows AI Overviews cite an underlying source around 68% of the time, favouring clean, extractable passages. Beyond that, publish things only you can publish: original data, real prices, genuine case studies, specific local knowledge. AI systems reward primary sources because everyone else is producing the same recycled summaries. And since ChatGPT alone accounts for roughly three-quarters of all AI referral traffic today, being visible and citable across the wider web - not just on Google - is now part of the job.
What this means for the site you build
This is where the plumbing of your website stops being a technical afterthought. An AI can only cite what it can cleanly read. Sites built on bloated page-builders, with content trapped inside images or rendered late by heavy JavaScript, are effectively invisible to the crawlers that feed these systems. The work we do - fast, clean, semantically structured builds with proper schema baked in from the start - happens to be exactly what makes a site legible to AI. When we build a site now, we treat machine readers as a first-class audience alongside human ones: descriptive headings, sensible markup, a clear content hierarchy, and structured data on every service and location page. It is the same craftsmanship good developers have always valued, pointed at a new and rather demanding reader. If your current site was built for the search era of 2020, it is quite possibly leaving you out of the conversations happening in 2026. We cover more of the practical side of this in our Insights.
If you take one thing from this, make it an audit rather than a panic. Look at your analytics for referrals from ChatGPT, Perplexity, Gemini and Google's AI surfaces - most businesses have never checked, and the numbers are often already sitting there. Then ask a plain question of your own site: if an AI read this page cold, could it explain what you do and quote a clear answer back to a customer? If the answer is no, the fix is rarely a rewrite of everything; it is usually cleaner structure, proper schema, and a handful of pages rebuilt to answer the questions your customers actually ask. That is a focused weekend of work for a capable developer, not a six-month project, and it compounds quietly every month the shift continues. You can see how we approach that kind of build on our services, and if you'd like a straight answer on what it would take for your site, we can put a number on it in minutes.