Total visits to the world's top 1 million websites fell 12.2% between their peak in the first quarter of 2022 and the third quarter of 2026, according to research SparkToro published on November 10, 2026. The analysis, written by SparkToro co-founder and chief executive Rand Fishkin, draws on Similarweb's global desktop and mobile clickstream panel and covers 20 quarters, from the fourth quarter of 2021 to the third quarter of 2026.

The headline figure sits beside a second, larger one. Over roughly the same period, the number of people online grew by about 24%, according to DataReportal estimates reproduced in the post. Divide falling visits by rising users and the result, by SparkToro's calculation, is a 29.2% drop in visits per web user. That second number rests on a population series whose own publisher has said it jumped in 2025 mainly because official figures for India and China were revised - a point the post does not address.

In Short

SparkToro and Similarweb data show visits to the top million websites fell 12.2% from their early 2022 peak, while the number of people online kept growing. That matters to publishers, SEO teams and buyers of open web inventory, because the pool of browser visits to share out has shrunk rather than grown. The biggest 100 sites gained only about 3 percentage points of share, so the decline is spread across the web rather than concentrated on small sites. The 29.2% fall in visits per user leans on DataReportal user counts that DataReportal itself says rose in 2025 largely because of revised official figures for India and China, so the per-person decline is probably smaller than the headline suggests.

How the research was built

Fishkin asked Similarweb's research team for two sets of numbers, according to the post. The first was the total traffic flowing to the top 1 million sites each month from 2021 to 2026. The second was a breakdown of that traffic across five rank bands: the 10 most visited domains, ranks 11 to 100, ranks 101 to 5,000, ranks 5,001 to 100,000, and ranks 100,001 to 1,000,000. Each band came with over-time metrics, so the trend could be followed quarter by quarter.

The charts carry the source line "Similarweb's global desktop & mobile clickstream panel (2021-2026)." Similarweb's Sam Sheridan and Adelle Kehoe are credited with helping construct the report, providing the data and validating the analysis. The population figures come from DataReportal's quarterly reports, which the post lists in a table running from 4.95 billion in the fourth quarter of 2021 to 6.15 billion in the third quarter of 2026.

The traffic totals are modelled estimates, not server counts. "Similarweb's global panel models traffic based on all those users," the post says, referring to the growing population of web users. The research question started as a hunch. Fishkin wrote that the project began with a theory that the top few websites, "dominated by the Big Tech players," were hoarding more of the web's traffic while fewer visits reached small and mid-size sites.

The method has a hard boundary. "We're only able to observe mobile and desktop web traffic via the browser," Fishkin wrote. "In the app world, consolidation and domination by Big Tech is likely far more prevalent." Activity inside apps such as Instagram, TikTok or a banking app does not appear in these figures at all. The analysis is also global, with no regional breakdown.

The giants' share barely moved

The first finding ran against Fishkin's own expectation. An annotation on one of the post's charts says so directly: "I expected the top 10-100 sites would have a much higher share of traffic in 2026 than 2021." They did not, by much.

The top 10 domains took 22.5% of visits to the top million in the fourth quarter of 2021 and 24.4% in the third quarter of 2026, according to the Similarweb chart. Sites ranked 11 to 100 moved from 10.8% to 11.8%. Together, the top 100 went from 33.3% to 36.2%, a gain of 2.9 percentage points - the "slight uptick of ~3%" Fishkin describes.

The three lower bands absorbed the difference. Ranks 101 to 5,000 slipped from 26.3% to 25.9%. Ranks 5,001 to 100,000 fell from 26.7% to 25.1%. The long tail, ranks 100,001 to 1,000,000, went from 13.7% to 12.7%, after peaking at 15.1% in the fourth quarter of 2023 and bottoming at 11.5% in the second quarter of 2025.

"First, the good news: the distribution of traffic to the open web has not changed much in the last 5 years," Fishkin wrote. "Despite my fears that Big Tech's monopolies were stifling traffic to the other players in the top million, it appears that stagnation is a more accurate description of the trend."

The quarterly path is less smooth than the endpoints suggest. For more than three years, the top 10 share drifted between 21.5% and 22.5%. Then, between the first and second quarters of 2025, it jumped from 22.5% to 24.1%, and the top 100 combined rose from 34.2% to 36.2%. It has stayed in that higher range since, with the top 10 reaching 24.8% in the second quarter of 2026. More than two-thirds of the five-year concentration gain therefore arrived in a single quarter. The post does not discuss that step or what might have caused it.

One sentence in the post contains an apparent slip. It compares the top 100 sites with "the following 9,999,900 domains," but the dataset covers the top 1 million, which leaves 999,900 domains below the top 100.

Smaller shares of a smaller total

Shares tell only half the story, because the total shrank. Applying each band's share to the totals the post reports gives a rough picture of absolute visits. These are PPC Land calculations from rounded chart percentages, not figures published by SparkToro or Similarweb.

On that basis, the top 100 sites received about 740 billion visits in the first quarter of 2022 and about 713 billion in the third quarter of 2026, a decline of roughly 4%. Sites ranked 101 to 1,000,000 fell from about 1.5 trillion visits to about 1.26 trillion, down roughly 17%. Within that group, the long tail of ranks 100,001 to 1,000,000 dropped from about 307 billion to about 250 billion, close to 19%.

Every band, in other words, received fewer browser visits in the third quarter of 2026 than at the 2022 peak. The loss was simply several times larger further down the rankings. "Stagnation" describes the shares; the volumes moved down across the board.

Fewer visits, more people

The second finding is the one Fishkin called less happy. Total traffic to the top million sites peaked in the first quarter of 2022 at 2.243 trillion visits, according to the post's summary. It reached its low in the first quarter of 2026 at 1.914 trillion, 14.7% below the peak. In the third quarter of 2026, the latest quarter in the data, it stood at 1.970 trillion, 12.2% below the 2022 high.

The decline was not a steady slide. On the chart, the bars sit around 2.2 trillion through 2022 and 2023, step down to around 2.1 trillion through 2024, and fall below 2 trillion from the second quarter of 2025 - the same quarter in which the top 10's share jumped. The third quarter of 2026 figure sits about 2.9% above the first-quarter low, by PPC Land's calculation. The post does not describe that as a recovery.

Meanwhile, the online population kept growing. DataReportal's figures, as tabulated in the post, rise from 4.95 billion in the fourth quarter of 2021 to 5.16 billion a year later, 5.35 billion at the end of 2023, 5.56 billion at the end of 2024, 6.12 billion in the first quarter of 2026 and 6.15 billion in the third quarter of 2026. The table lists no report for the fourth quarter of 2025 or the second quarter of 2026, though the population line on the chart runs through both quarters.

"Astute readers will have already caught the problem - there's 25% more people using the web, and a decline in total traffic," Fishkin wrote. The post's summary puts the growth at about 24%, from 4.95 billion to 6.15 billion, and states that visits per user "fell more than twice as much as total traffic." Its figure: a drop of 29.2%.

The numbers are not perfectly aligned across the post. The body text says total visits to the top million "have fallen ~10%," while the summary and headline use 12.2%. The summary labels the 4.95 billion user figure as 2022, while the table assigns it to the fourth quarter of 2021. The 29.2% figure matches a fourth-quarter 2021 baseline. Using the first quarter of 2022 for both traffic and users - 2.243 trillion visits against 5.00 billion people - gives about 449 visits per estimated user, against about 320 in the third quarter of 2026, a decline of 28.6% by PPC Land's calculation. The post does not state whether each quarterly bar is a three-month sum or a monthly average, which affects the per-user level but not the percentage change.

The population problem

A per-user figure is only as reliable as its denominator, and this denominator has a known break.

The post's table shows 5.65 billion web users in the second quarter of 2025 and 6.04 billion in the third quarter of 2025 - an increase of 390 million, or 6.9%, in one quarter. That single step accounts for about a third of the 1.2 billion increase across the whole period, by PPC Land's calculation.

DataReportal has explained that step itself. In its report on internet users passing 6 billion, published on October 15, 2025, DataReportal attributed most of the apparent jump to significant revisions in internet adoption rates reported by official bodies in India and China. According to that report, the change should not be read as hundreds of millions of people coming online within 90 days. DataReportal's notes on its data add that the revisions may have an important effect on figures for change over time, and urge caution when comparing data from different reports. The same release cycle produced the Digital 2026 figures of 5.66 billion social media user identities that PPC Land covered that day.

The effect on SparkToro's headline per-user number is measurable. If the 390 million step were removed entirely, users in the third quarter of 2026 would be about 5.76 billion, growth since the fourth quarter of 2021 would be about 16% rather than 24%, and the fall in visits per user would be about 24% rather than 29.2%. Those are PPC Land calculations, and they mark an outer bound: part of that quarter's increase may reflect genuine growth, so the true figure probably sits between the two. The direction holds either way. The size does not.

There is a second dependency the post leaves open. If Similarweb's model scales its panel to an estimate of the online population, a revision to that estimate could feed into the traffic totals as well. The post does not say which population figures Similarweb uses, or whether its model absorbed the 2025 revisions.

What the data does not establish

The post is candid about the limits of its own conclusions. Fishkin lists four possible explanations without choosing between them: more people reaching the internet exclusively through apps; search engines, AI tools and social networks keeping people from needing, or wanting, to visit more websites; the waning effect of the Covid-19 pandemic, since "2021-2022 had especially high Internet usage during the lockdowns"; or another confounding factor.

"We can't say for certain," Fishkin wrote. "But we know that total web traffic and visits-per-Internet-connected human are down from their peak in 2021/22."

The baseline matters. A series that begins during lockdown-era usage will show declines that partly reflect a return to normal behaviour, and the data does not separate that effect from any structural shift. Nor does it show where the lost visits went - to apps, to AI assistants, to search result pages, or nowhere - because only browser visits to websites are measured.

Geography is another blind spot. The post acknowledges it is "plausible that we've seen a decline in traffic to US/UK/EU websites (as many have reported) while other regions have grown with increasing Internet access," citing Asia and Africa. A worldwide total can hide sharper falls in mature markets and gains in growing ones. The research offers no regional split to test either possibility.

Human panels and server logs also see different webs. Cloudflare Radar data for the seven days ending June 5, 2026 showed bots accounting for 57.4% of traffic to HTML content, against 42.6% for humans. A clickstream panel records people; a publisher's own logs record everything that requests a page. The two will not move in step.

Finally, the research does not claim to mark the start of the shift. "2021 was hardly the start of the Zero-Click era," the post says, pointing to SparkToro's earlier research, which documented changes in Google search clicks and social platforms biasing against links between 2012 and 2020. Fishkin wrote that he hopes to validate the findings with additional sources and to watch the numbers "every year or two."

How it fits with earlier measurements

SparkToro has published research on outbound traffic from large platforms for several years, and PPC Land has covered much of it.

In March 2024, Datos data analysed by SparkToro showed Google accounting for 63.41% of all US website referrals, with referrals to sites outside the top 170 shrinking. In July 2024, a SparkToro study of more than 1.6 million Google searches found that only 374 of every 1,000 US searches produced a click to the open web, against 360 per 1,000 in the EU. In May 2025, Fishkin described seven years of decline in Google's organic click-through rates, arguing that AI tools and social platforms pass on very little of the traffic they receive.

Datos panel data, which covers desktop users in the US, EU and UK, put the share of US zero-click searches at nearly 27% by September 2025. By the first quarter of 2026, AI tools accounted for 1.65% of US desktop events, up from 1.31% a year earlier - still a small slice. Definitions vary widely. NPR's Planet Money put the share of US Google searches ending without a click at 68% in July 2026, a figure PPC Land set against Datos and Similarweb numbers when covering All About Berlin's 75% traffic loss.

Fishkin and Amanda Natividad turned the subject into a book. When PPC Land covered the Zero Click Marketing pre-orders in May 2026, the 46% web traffic decline in the headline came from a LinkedIn post that gave no method or dataset. The SparkToro blog listed the book as live on November 1, 2026. The new research measures something narrower and more precisely defined - worldwide browser visits to a fixed set of a million sites - and produces a far smaller decline.

Publisher-level data points the same way, though with larger numbers. Similarweb figures published by Press Gazette showed 44 of the top 50 US news sites with fewer visits in July 2025 than a year earlier. A Bloomberg investigation using Similarweb data for 67 small publisher sites recorded declines of 70% or more at several publishers. Ahrefs estimated in April 2025 that AI Overviews were associated with a 34.5% lower click-through rate for top-ranking pages, a figure based on a forecast of what the click-through rate would have been without the feature. On November 1, 2026, PPC Land reported that Digital Content Next had told a federal court its members' unique visitors fell 26% between 2023 and 2025, without describing its survey method. JWX's claim in August 2026 that search referral had fallen from up to 70% of publisher visits to under 20% came with no dataset at all.

Against those figures, a 12.2% decline across the top million looks modest. That gap is not a contradiction. Individual publishers and specific referral channels can fall much faster than an aggregate that also includes Google, YouTube, Wikipedia and Amazon. The SparkToro data shows the aggregate; it cannot show which sites inside each band gained or lost.

Why it matters for the marketing community

For publishers, the arithmetic behind advertising revenue starts with visits. A smaller pool of browser visits, spread among roughly the same distribution of sites, means less inventory to sell across the open web as a whole, even before pricing is considered. The SparkToro data says nothing about ad spend, CPMs or impressions, so the revenue effect cannot be read off the chart.

For search marketers, the distribution finding cuts against a widely held assumption: that the largest platforms are absorbing an ever-larger share of web visits. In browser traffic, at least, the top 100 sites moved less than 3 percentage points in five years, and most of that in one quarter of 2025. Where concentration may have grown faster, Fishkin suggests, is inside apps - the walled gardens this dataset cannot see.

For buyers of open web media, the research arrives during an argument about the future of that inventory. On October 27, 2026, PPC Land covered an essay in which The Trade Desk's chief executive Jeff Green said Fortune 500 advertisers would decide the open internet's fate; the essay cited no figures on ad spend or time spent. SparkToro and Similarweb supply one measure of the underlying audience: fewer visits than four years ago, from more people, spread across the web in roughly the same proportions.

What the data cannot yet say is whether the visits that disappeared moved somewhere advertisers can still reach, or simply stopped happening. SparkToro plans to revisit the numbers. On current evidence, the per-user decline is real but probably smaller than 29.2%, the total decline is about 12%, and the cause remains unproven.

Timeline

Summary

Who: SparkToro, through co-founder and chief executive Rand Fishkin, using clickstream panel data from Similarweb, with Similarweb's Sam Sheridan and Adelle Kehoe credited for the data and validation, and DataReportal's quarterly estimates of global internet users.

What: Total visits to the top 1 million websites fell 12.2% from a peak of 2.243 trillion in the first quarter of 2022 to 1.970 trillion in the third quarter of 2026, while estimated web users rose about 24%, giving a 29.2% drop in visits per user by SparkToro's calculation. The share of visits going to the top 100 sites rose only from 33.3% to 36.2%. About a third of the user growth comes from a single 2025 quarter that DataReportal attributes largely to revised official data for India and China; removing it entirely would cut the per-user decline to about 24%, by PPC Land's calculation.

When: SparkToro published the research on November 10, 2026. The data covers the fourth quarter of 2021 to the third quarter of 2026.

Where: Worldwide browser traffic on desktop and mobile, with no regional breakdown and no app usage.

Why: The analysis began as a test of whether Big Tech's largest sites were taking a growing share of web traffic. The distribution barely moved, but the overall pool of browser visits shrank as the online population grew. SparkToro does not identify a cause, listing app usage, search engines, AI tools and social networks, the end of pandemic-era usage, and other factors as possibilities. For publishers, search marketers and open web advertisers, the figures describe a smaller audience pool, though the data says nothing about ad spend or revenue.