The number of people using the internet grew by roughly a quarter in five years. The number of visits to the open web stayed almost exactly the same.
Sit with that for a second. More than a billion new people came online, and the websites did not notice.
The data behind the flat line
Rand Fishkin published the analysis on the SparkToro blog this week, built with Similarweb's research team. He asked for the total monthly traffic flowing to the top one million websites from 2021 to 2026, split into buckets: the top 10 domains, positions 11 to 100, 101 to 5,000, and so on down to the millionth site.
His starting theory was that Big Tech was hoarding a growing share of traffic. The data said no. The distribution between the giants and everyone else barely changed in five years.
The second finding was the uncomfortable one. DataReportal estimated about 4.95 billion internet users at the end of 2021 and about 6.15 billion by the third quarter of 2026. Similarweb's panel models traffic across all of those users, and the total stayed flat. Visits per connected person fell by roughly a quarter.
Fishkin is careful about the caveats, and they matter. The panel only sees browser traffic, not apps. It is a worldwide view, so declines in the US and Europe can be masked by growth in Asia and Africa. And the zero-click shift started long before 2021.
Where the visits went instead
The study does not say why. Fishkin lists the candidates: more people living inside apps, search engines and AI assistants answering questions directly, social platforms that punish outbound links.
My read is that all three are true at once, and the cause matters less than the operational reality. The new internet user of 2026 often arrives through a phone, inside an app, with an AI assistant one tap away. That person was never going to browse ten websites to compare vendors. They ask, they get a shortlist, and they visit one or two sites at the end.
For a company, the consequence is simple. Your addressable audience grew. Your addressable traffic did not. The research still happens, just not on your pages.
Why the flat line is worse than a decline
A falling traffic chart gets attention in a board meeting. A flat one does not. Flat looks like stability, and stability looks like the plan is working.
But flat traffic in a market that grew 25% means your share of attention shrank. If your category gained buyers and your site did not gain visits, the buyers found answers somewhere else. Someone else is shaping their first impression of you.
I see this pattern constantly in audits we run at difrnt. Organic sessions hold steady year on year, so the SEO report says green.
Then you look at branded search volume, sales cycle length and the share of leads who already have a shortlist when they call, and the story changes. The website stopped being where the decision starts. It became where the decision gets confirmed.
What this does to the marketing plan
The first adjustment is in forecasting. Most traffic targets are set as last year plus a growth percentage. That model assumes visits scale with audience. The SparkToro data says they no longer do, so a traffic target can be hit while the business loses ground.
The second adjustment is in where content lives. If each person visits fewer sites, the content that wins is the content that travels without a click. Useful posts on LinkedIn that stand on their own, videos that answer the question in the first ten seconds, data that other people cite, and pages written so AI assistants can extract a clean answer.
I wrote about this in your website is the source, not the destination. The SparkToro numbers make that argument with cold data instead of opinion.
The third adjustment is in measurement. If the visit is the last step instead of the first, then measuring only visits measures only the end of the journey. You need signals for the middle: branded search, direct traffic, mentions in AI answers, share of voice in the places your buyers actually spend time. If you want to see how your brand shows up in AI-generated answers specifically, that is the problem GEOflux.ai was built to measure.
The part nobody wants to say out loud
For fifteen years, digital marketing ran on a quiet assumption. More people online meant more potential visitors, and more visitors meant more pipeline. Growth in internet adoption was free tailwind for every website.
That tailwind stopped. The pie of people grew and the pie of clicks did not. Visits are now closer to a zero-sum game, where one site's gain is someone else's loss, instead of a market that grows on its own.
This is not a reason to abandon the website. It is still where trust gets confirmed, where pricing lives, where conversion happens. It is a reason to stop treating traffic as the scoreboard and start treating it as one input among several.
Fishkin says he plans to rerun the numbers every year or two. I hope he does, because this is the kind of baseline most marketing teams never measure and almost everyone assumes.
The audience is still growing. The visits are not. Plan for the gap.