Why every zero-click figure is different#
The number quoted most right now is 68%. It comes from SparkToro's analysis of Similarweb data for January to April 2026: 68.01% of US Google searches ended with no click at all, up from 60.45% in 2024 (Search Engine Land). It assumes two thirds of searches happen on a phone, and it leaves out searches made inside the Google app.
The biggest dataset I've seen comes in lower. iPullRank analysed 10.2 billion Google search sessions from the Datos panel, October 2024 to December 2025, and found 46.96% ended without a click (iPullRank). But 94.5% of that panel was desktop, the device where people click the most.
The same study has the figure I'd pay most attention to as a UK business: 65% of UK mobile searches ended without a click, against 46% on desktop. That 19-point gap was the largest of any country it looked at.
Three things push these numbers around: the mix of phones and computers, whether a tap through to Google Maps or YouTube counts as a click, and which searches were sampled in the first place. None of the studies is wrong. None of them tells you what is happening on your own site either.
What I take from them is simple. If your customers search on their phones, and for most local and consumer businesses they do, a large share of their searches now finish on Google.
People haven't stopped using Google#
Falling clicks get read as people abandoning Google for ChatGPT. The evidence doesn't support that, at least not yet.
Google said in March 2025 that it handles more than 5 trillion searches a year (Search Engine Land). In August 2025 its Head of Search, Liz Reid, wrote that total organic clicks from Google to websites had been "relatively stable" year on year (Google). Google didn't publish the data behind that, so treat it as a claim rather than a measurement.
Independent data points the same way. Semrush tracked US desktop users before and after they started using ChatGPT, across 260 billion rows of clickstream data, and found no meaningful drop in how often they used Google afterwards (Semrush).
AI assistants do send visitors, just not many of them. The SEO Works pooled GA4 data from 95 client websites for the twelve months to August 2026. Organic search delivered 39 times as many sessions as all AI assistants combined, and by August AI referrals had only reached 3.86% of organic volume (The SEO Works). The AI visitors converted slightly better (7.18% against 6.08%), but they produced roughly 3 conversions for every 100 that came from organic search.
So track your AI referrals, by all means. Just don't rebuild your marketing around them. Google is still where most of your search customers start.
Which searches still send people to your website#
The pattern is easy to spot once you know what to look for. A search ends on your site when the person has to do something Google can't finish for them: buy, book, get a quote, check stock, log in, or read the detail they need before spending money. It ends on Google when a sentence or a map listing is enough.
Ahrefs measured how often AI Overviews appear across 146 million desktop results in September 2025 (Ahrefs). An Overview showing up doesn't automatically cost you the click, but it shows where Google has decided to do the answering itself:
| Type of search | Example | AI Overview appeared on | What it means for you |
|---|---|---|---|
| Your business name | "Smiths Joinery kitchen quote" | 13.1% of branded searches (0.9% of purely navigational ones) | Safe when they want to buy, book or contact you |
| Ready to buy | "oak dining table 6 seater" | 2.1% of transactional searches, 3.2% of shopping | Mostly still clicks, though Google Shopping and ads compete for them |
| Local | "emergency electrician Solihull" | 7.9% of local searches | Plenty of calls and directions, fewer website visits |
| A question | "how long does a boiler service take" | 57.9% of question searches | Often answered on Google |
| Health, money, legal | "can I claim for a delayed flight" | 34.3% of YMYL searches (health, money, legal), 44.1% of medical | Explainers heavily answered; local and calculator searches far less |
If most of your website traffic comes from the bottom two rows, you're exposed. If it comes from the top three, zero-click mostly happens to somebody else, although you may still see it in your analytics (more on that further down).
Local businesses face two different searches#
This distinction matters more for local businesses than anyone. Whitespark checked 540 local searches across six industries in three US cities. AI Overviews appeared on 15% of searches looking for a provider ("personal injury lawyers in Phoenix"), but on 92% of local questions and 97% of mixed ones ("should I get a lawyer after a car accident in Phoenix?"). The map pack did the opposite: it showed on 93% of provider searches and just 6% of the questions (Whitespark).
In practice, a roofer keeps "roofers in Solihull" and loses "how long should a flat roof last". The searches that bring the work in are still there. The ones that used to bring blog traffic are going.
There's a twist, though. Winning the map pack often produces no website visit at all. The customer calls straight from your listing, asks for directions, or reads your reviews and turns up. That's a zero-click search that still pays. Your website analytics won't show it, but your Google Business Profile performance report will: calls, direction requests, bookings and website clicks. Judge local SEO on website sessions alone and you'll undercount it every time.
Those were US cities and I haven't seen an equivalent UK study. The mechanics are the same here, so I'd trust the direction and treat the exact percentages as indicative.
No click doesn't always mean no customer#
Some people genuinely stop at the answer. Pew Research tracked the Google searches of 900 US adults in March 2025. When an AI summary appeared, they clicked a normal result on 8% of searches, against 15% when it didn't. They clicked a link inside the summary only 1% of the time, and ended their browsing altogether after 26% of those searches, against 16% (Pew Research Center).
Others come back later by a different route. Similarweb followed US desktop users after ChatGPT recommended a brand they hadn't named in their question or visited in the previous four weeks. The recommended brand was 2.5 times more likely to get a visit within seven days than a competitor that wasn't mentioned, and most of those visits arrived through an ordinary search rather than a trackable AI link (Salience, summarising Similarweb's study). The brands were large finance, travel and beauty names, so whether the effect holds for small businesses is still unproven.
The practical problem is attribution. Someone who saw your name in an answer and searched for you the next day shows up as branded organic traffic. Someone who typed your web address shows up as direct traffic rather than organic. Neither gets credited to the search that actually introduced you.
The cheapest fix is to ask. Add "How did you hear about us?" to your enquiry form and booking process, with ChatGPT and "Google's AI answer" listed alongside the usual options.
Which small businesses are most exposed#
How exposed you are has little to do with your size. It depends on what kind of searches bring your visitors in.
Most exposed: Businesses whose traffic comes mainly from answering questions. Content-led sites, small publishers, affiliates, and the guides section of a professional services site. If a paragraph gives someone what they wanted, Google can serve them that paragraph.
In the middle: Professional services and B2B. Discovery content ("what is…", "do I need…") loses clicks. The decision stage doesn't, because nobody hires an accountant or signs a software contract off a summary. Expect fewer early visits, from people who are further along when they do arrive.
Least exposed: Local trades, hospitality and e-commerce, where the customer has to call, book, order or turn up. Your analytics might still show a dip as more of those actions happen inside Google, so check what happened to enquiries before you panic.
The clearest illustration of how uneven this is comes from Germany. SISTRIX analysed more than 100 million German keywords and found that where an AI Overview appeared, the top result's click-through rate fell from 27% to 11%. Across the whole market, though, organic clicks dropped by about 6.6%, because only around a fifth of keywords showed an Overview. Booking.com lost 0.46% of its organic clicks and Amazon 1.73%. Specialist health sites lost 29 to 30%, and parenting sites more than 24% (Search Engine Journal).
Same feature, same country, and losses ranging from under half a percent to nearly a third. An average tells you almost nothing about your own site.
Before you blame AI for a traffic drop#
Zero-click is the easiest explanation to reach for right now, and it's often the wrong one. Put three numbers side by side for the same dates:
- Search Console clicks (what Google says it sent you)
- Analytics sessions (what your site recorded)
- Enquiries, calls, bookings or orders (what actually happened)
Then match the pattern:
Impressions and average position steady, clicks down, mostly on question-type searches: Likely zero-click.
Positions down, starting during a Google core update: A ranking change. Google confirmed core updates in March, June and December 2025, and in March and May 2026.
Impressions down, position steady, and the same dip this time last year: Seasonal or falling demand. Google Trends will confirm it.
A cliff on the day you launched a redesign or moved platform: Technical. Redirects, indexing and tracking are the usual suspects.
Search Console steady but analytics down sharply: Your tracking, not your customers. UK sites are particularly prone to this when the cookie banner doesn't pass consent choices through to Google's tags.
I go through each of these in more depth in why your organic traffic is dropping. If you have Search Console access and want to see exactly which pages are affected, I walk through how to check your site for AI Overview losses step by step.
What to measure instead of website traffic#
Sessions still matter. They just can't be the whole picture any more. For a small business I'd track five things:
Treat that last one as a spot check, not a ranking. AI answers change with the person asking, their location and the day, so look for a direction over several months rather than reading anything into a single result.
Beyond these, the SEO metrics that matter covers the numbers that actually predict revenue.
Three mistakes I'd avoid#
Giving up on SEO. In SEOFOMO's 2025 survey of 211 search professionals, 62% put AI search at 0 to 5% of their sites' revenue, and another 30% couldn't say (SEOFOMO). For most businesses, Google search is still where the money comes from. Pulling budget away from it because of zero-click headlines is a bigger risk than zero-click itself.
Paying for AI visibility scores as if they were rankings. These tools run a sample of prompts and report a percentage. That's useful for spotting a trend, but answers vary by user, location and model version, so there's no stable position to climb. Before you pay, ask the vendor which prompts they test, how often they repeat them, and whether you can see the raw answers.
Publishing more generic content to "feed the AI". The content losing the most clicks is the content anyone could have written, so producing more of it makes the problem bigger. What still earns the visit is what a summary can't reproduce: your prices, your availability, your process, your photos, your results, and opinions grounded in real experience. It's the same reason most SEO advice doesn't apply to your site until you adapt it to what you actually sell.
The short version: Zero-click is real, it's growing fastest on mobile, and it lands very unevenly. Work out which of your searches are questions and which are actions, measure what customers do rather than only what your site records, and the headline numbers lose most of their sting.