Same Google Position, Ten Times the Clicks: What 40% Publisher Losses and Our Own Search Console Data Say About 2026

Publishers lost around 40% of their Google traffic in a year. That headline has been repeated enough that it has stopped meaning anything to a plumber, a florist or a shop owner. So we checked it against something we can actually see: our own Search Console data across three UK sites.

What we found is more useful than the headline. The clicks that are disappearing are not evenly spread. Searches that want an answer are losing clicks fast. Searches that want to buy, hire or find one specific business are holding up. That difference is the strategy.

The short version

• Chartbeat reports that Google Search referrals to publishers fell 40.2% between July 2025 and July 2026, and small publishers have been hit hardest.
• Pew Research found people clicked a normal result 8% of the time when an AI summary appeared, against 15% when it didn't.
• In our own data, a buying-intent query at position 7 earned a 14% click-through rate. An informational query at the same position earned 1.4%.
• Google's own 15 October 2026 changes put more links into AI answers. That tells you where it thinks the pressure is.
• For local and ecommerce businesses the practical answer is not "write less". It is to put buying-intent pages first and keep informational content fresh, sourced and original.

What companies are actually seeing

Start with the widest data set. Chartbeat's 2026 Publisher Playbook, covering mostly news and media sites, shows Google Search referrals down 40.2% year on year to July 2026, after a 21.9% fall the year before. Google Discover is down 34.3%. Google Search's share of publisher pageviews has fallen from about 9% to about 5% in two years.

What companies are seeingInteractive

Who lost Google traffic, and who didn’t

Switch between publisher size, named companies and where the traffic went instead.

By publisher sizeNamed companiesWhere traffic moved

Small publishers

~1k–10k page views a day

−60%

Mid-sized publishers

10k–100k a day

−47%

Large publishers

100k+ a day

−22%

−40.2%

Google Search referrals to publishers, July 2025 to July 2026 (Chartbeat)

−34.3%

Google Discover referrals over the same year (Chartbeat)

0.01%

Share of publisher traffic from AI referrals, up eightfold in two years (Chartbeat)

The split by size matters more than the average. Chartbeat data reported by Axios in March put small publishers (roughly 1,000 to 10,000 page views a day) at a 60% two-year loss, mid-sized at 47% and large at 22%. If you run a small business site, you are much closer to the first group than the last.

Where did the traffic go? Not to AI chatbots, at least not yet. Chartbeat says AI referrals have grown eightfold in two years but still make up about 0.01% of publisher traffic. The share that grew was direct visits and dark social, which suggests people typing a name, using a bookmark or clicking a link shared privately. In other words, people who already know who they want.

Big names are reacting. In July the WSJ reported that USA Today, Politico, The Economist, People Inc., Reuters and Reddit were weighing blocking Google. People Inc.'s CEO said full blocking is "on the table", and the company says Google supplied 25% of its traffic in the first quarter of 2026, down from over half two years earlier. Treat this as leverage in a negotiation rather than a plan: blocking Google's crawler also removes a site from AI answers, and one study found publishers that blocked AI crawlers lost between roughly 7% and 23% of visits.

It is not a uniform collapse. Semrush estimates reported in the same coverage suggested the Guardian and the BBC were up year on year while the Washington Post and Reuters were down by a third or more. Those figures were corrected after publication, so read them as direction, not precision.

How people search now

Three data points describe the behaviour change:

1. Pew Research Center tracked the browsing of 900 US adults in March 2025. When an AI summary appeared, 8% clicked a traditional result; when it didn't, 15% did. Only 1% of visits with a summary clicked a link inside the summary itself.
2. SparkToro reported that 68% of US Google searches ended without a click in early 2026.
3. Google itself announced on 15 October 2026 that it is adding substantially more inline links to AI Mode and AI Overviews, showing them higher up, testing a section for "official" pages, and letting creators' names link to their profiles. It also said use of its Preferred Sources feature has grown more than seven times since January.

That last point is the quiet admission. Google is not adding links because it is feeling generous. It is adding them because the click economy that funds the web is under strain, and publishers are threatening to leave.

Our own data: the click gap

Here is the part you won't find in any industry report. We looked at Search Console for georanklocal.com for the 28 days to 27 September, and at two other UK sites in our case-study portfolio, padelchippenham.co.uk and sleepertrain.co.uk, for the three months to the same date.

Our own Search Console dataInteractive

Same Google position. Very different clicks.

Pick a real query from our Search Console exports. Each grid is 100 people who saw the result; the filled dots are the ones who clicked.

100 searchersBy position band

• “find questions”GeoRankLocal tool · position 6.81.37%
• “sleeper train liverpool”sleepertrain.co.uk · position 7.593.45%
• “how much is the caledonian sleeper”sleepertrain.co.uk · position 7.94.17%
• “train carriage for sale”sleepertrain.co.uk · position 7.1814.07%
• “railway carriage for sale”sleepertrain.co.uk · position 7.2810.81%
• “question finder”GeoRankLocal tool (own name) · position 2.7717.6%
• “padel chippenham”padelchippenham.co.uk · position 1.2668.76%
• “chippenham padel courts”padelchippenham.co.uk · position 1.0180.25%

Buying intent37 clicks from 263 impressions

14 in 100 clicked

Someone wants to buy or hire. They have to leave Google to do it.

Three things stand out.

Intent beats position. A query at position 7 about buying a railway carriage earned a click-through rate of 14.1%. A query at position 6.8 for finding questions earned 1.4%. Even within the same site (sleepertrain.co.uk), informational queries at position 7 to 8 ("sleeper train liverpool", "how much is the caledonian sleeper") earned 3.5% to 4.2%, a third or less of the buying queries' rate. Different sites, different windows, so it is an observation, not an experiment. But the gap is far too large to be noise.

Brand and local searches barely lose clicks. Padel Chippenham (padelchippenham.co.uk), a local venue in our case-study portfolio, got 810 clicks from 1,178 impressions on "padel chippenham" (68.8%) and 80% on "chippenham padel courts". When someone has already decided who they want, there is no summary to read instead of clicking.

We are our own cautionary tale. Of georanklocal.com's 125 clicks in those 28 days, 116 came from two free tools, and 33 came from India and 28 from Turkey. The UK produced 24% of our impressions but only 8 clicks, at an average position of 49. Our commercial pages are not ranking yet. And in August a spam update erased most of our impressions in a week, almost all from one free tool, which taught us the same lesson from the other side: traffic that doesn't buy is the first to go.

If that sounds like a confession, it is meant as one. The strategy we sell is the strategy our own numbers force on us.

What this costs you

Averages are only useful if you can translate them. Put your own numbers into this, using Search Console and your own enquiry rate. It is deliberately a what-if, not a forecast.

Illustration, not a forecastInteractive

What could this cost your business?

Plug in your own numbers from Search Console and pick the benchmark you think fits. The sums are shown so you can check them.

AI summary effectPublisher averageSmall-publisher worst case

Pew Research (900 US adults, March 2025): 8% clicked a traditional result when an AI summary appeared vs 15% when it didn’t, a relative drop of about 47%. Applied only to the share of your informational searches that show a summary.

Google clicks a month500Share that lands on informational pages60%Of those searches, share showing an AI summary40%Stress test: buying-intent clicks also fall by0%Enquiry rate (clicks that become enquiries)2%Average value of one enquiry£150

Clicks a month

444 left−56 lost

1.1

enquiries a month at risk

£168

a month in enquiry value

Your mix leans towards buying intent, which is the safer side. The risk is mostly in your guides and blog posts.

The main thing the calculator should show you is where your exposure sits. A tradesman whose clicks come from "emergency plumber near me" and a Google Business Profile has very little informational traffic to lose. A shop whose traffic comes from long "how to style linen" posts has a lot.

What's changing next, and how the big platforms are steering search

Nobody outside Google knows its roadmap, and we won't pretend to. What we can do is line up what has been announced and shipped this year.

TimelineInteractive

Ten months that changed how search sends traffic

Tap an event. Every item links to its source.

JanMarJulAug15 SepSep15 Oct

15 Oct 2026

Google puts more links into AI Mode and AI Overviews

More inline links, shown higher up. A test of a section highlighting one or two official pages. News carousels and clickable creator names. Preferred Sources use is up more than seven times since the start of the year. Google is responding to the click problem, but the links are still inside the answer.

Source: Google, The Keyword

Read together, three directions are clear enough to plan around.

Answers keep absorbing information-seeking searches. AI Mode and AI Overviews are being given more links, but the links sit inside the answer. Winning a citation inside the answer is becoming a second, separate job from ranking.

Buying is moving into the AI interface. Google's Universal Commerce Protocol, announced in January, powers a Buy button inside AI Mode and Gemini for select merchants in the US, Canada and Australia. OpenAI has a product feed specification and a checkout protocol of its own, again US-first. Neither is a UK reality today, but both tell you what "being found" is turning into: clean, current product and business data that a machine can act on.

Spam and quality systems are getting heavier. The September 2026 spam update ran for 13 days and 16 hours, the longest of any 2026 spam update, and Google says recovery can take months with no reconsideration request. Thin, off-topic or scaled content on a small site is a bigger risk than it was a year ago.

What "fresh" should mean now

Freshness gets treated as a date trick. Updating the date on an article does nothing for a reader, and we would not recommend relying on it for Google either. We can't see Google's ranking system, so what follows is our practice, not a rule.

• Update the content first, then the date. We show both the original date and an "Updated" date on every article.
• Lead with something new: a number, a dataset, a changed rule. This article leads with data from three sites you couldn't read anywhere else.
• Cite where every figure comes from and say which are estimates. Summaries and searchers both reward sources.
• For businesses with prices, stock or opening hours, freshness means the facts on the page, in the schema and in your feeds are identical and current. A stale "in stock" flag is worse than no data.

How exposed are you?

Self-checkInteractive

How exposed is your search traffic?

Seven yes/no questions. Nothing is stored or sent anywhere.

1. 1.Most of your Google clicks land on blog posts or how-to pages.

   Search Console, Pages report.

   YesNo

2. 2.One or two pages deliver more than half of your clicks.

   Our own site: two tools gave 116 of 125 clicks.

   YesNo

3. 3.Much of your search traffic comes from outside the area you serve.

   Search Console, Countries report.

   YesNo

4. 4.Your buying-intent pages (services, products) sit beyond page 2 of Google.

   Position above ~20 means almost no clicks.

   YesNo

5. 5.Your key pages have no structured data (LocalBusiness, Product, FAQ).

   Check with Google’s Rich Results Test.

   YesNo

6. 6.Your important pages haven’t had a real update in the last six months.

   Prices, opening hours, stock, stats.

   YesNo

7. 7.Few people search for your business name directly.

   Search Console, queries containing your name.

   YesNo

0 of 7 answered

Answer to see your score

0/100

Do this month

1. Open Search Console and split your queries by intent, not by total clicks. Informational, buying and brand/local each tell a different story.
2. Rank the money pages first. If your service or product pages sit beyond page two, nothing else on this list matters as much.
3. Turn each guide into a route to a buying page. Informational content should answer the question, then say what to do next and link to the page where you do it.
4. Publish one piece of original data a quarter. Your own enquiry numbers, prices, job times or customer questions are more citable than another summary of what everyone already says.
5. Fix the machine layer. Structured data that matches the page, a complete Google Business Profile, and (for shops) a clean product feed. Our ecommerce setup guide covers the shop side.
6. Measure branded and direct demand. It is the only click type that every source above shows growing.

Method and limits

• georanklocal.com figures are from Search Console exports for 31 August to 27 September 2026. The other two sites' figures are from three-month exports to 27 September. Windows differ and the sites are not the same.
• Search Console hides rare queries, so the band totals count only queries Google lists. The top-three band for georanklocal.com includes queries that contain our own tool names, which is brand search.
• Pew's sample is US adults in 2025; Chartbeat's clients are mostly news publishers. They are the best public data on click behaviour, but they are not a measure of your market.
• Press-reported figures, including Semrush estimates and the WSJ-sourced People Inc. data, were gathered by trade publications and some were corrected after publication. Check the sources before quoting them.

Sources

• Chartbeat 2026 Publisher Playbook, via Relevant Audience
• Small publishers' search referral traffic down 60%, Search Engine Journal
• Reddit and USA Today face Google exit as search traffic drops, PPC Land
• Pew Research on AI summaries and clicks, summarised by ODSC
• 68% of US Google searches ended without a click in early 2026, SparkToro, via TechWyse
• Google: AI Mode and AI Overviews updates to explore the web, 15 October 2026
• Google Merchant Center Help: Universal Commerce Protocol
• September 2026 spam update complete, Relevant Audience
• August 2026 spam update, Search Engine Roundtable