SparkToro Says Web Visits per User Fell 29%, So Grade Your Traffic on a Curve

SparkToro Says Web Visits per User Fell 29%, So Grade Your Traffic on a Curve
Open-web visits are down 12.2% from the 2022 peak while web users grew about 24%, per SparkToro and Similarweb.

SparkToro and Similarweb reported on November 10, 2026 that visits to the top 1 million websites fell 12.2% from their Q1 2022 peak, to 1.970 trillion in Q3 2026, while estimated web users grew about 24%. That works out to a 29.2% drop in visits per user. A site with flat traffic since 2022 is probably beating the open web, not stalling.

I think that second sentence matters more than the headline number. Most of the traffic reviews I see in marketing decks still treat the internet as a fixed-size pond. Last year minus this year, red arrow, someone gets asked to explain it. Rand Fishkin's new SparkToro research is a decent reason to stop doing that, or at least to put a second line on the chart.

What the Similarweb panel actually measured

The data comes from Similarweb's desktop and mobile clickstream panel, worldwide, Q4 2021 through Q3 2026. Browser traffic only. No apps. Total visits to the top million sites peaked at 2.243 trillion in Q1 2022, bottomed at 1.914 trillion in Q1 2026 (down 14.7%), and recovered a little to 1.970 trillion by Q3.

The surprise, at least to Fishkin, was the distribution. He went in expecting the biggest sites to have swallowed the web. They mostly didn't. According to PPC Land's breakdown of the rank bands, the top 100 sites went from 33.3% of top-million visits to 36.2%, and more than two-thirds of that gain landed in a single quarter (Q1 to Q2 2025). Sites ranked 100,001 to 1,000,000 slipped from 13.7% to 12.7%.

Put plainly: the pie shrank, and nearly everyone's slice shrank with it. PPC Land's own math from the rounded shares has the top 100 down roughly 4% in absolute visits, while the long tail (ranks 100,001 to 1M) fell roughly 19%. So the big sites did better, just not by the margin most of us assumed.

Fishkin is careful not to name a cause. His list of suspects includes apps, search and AI answers keeping people on the results page, social platforms, and plain post-COVID normalization. "We can't say for certain," he writes. Fair enough. And honestly, the lockdown baseline is the one I'd weight more than people want to, since 2021 and early 2022 were strange years for screen time.

The 24% user growth needs an asterisk

This is where I'd slow down before anyone pastes "visits per user down 30%" into a board slide. The user count comes from DataReportal, and a big chunk of the growth is a single step: roughly 390 million users added between Q2 and Q3 2025. PPC Land notes DataReportal attributes that jump largely to revised official figures for India and China, not 390 million people coming online in one quarter. DataReportal's own "six billion internet users" report mentions the revised China number specifically.

Strip that step out and PPC Land estimates user growth closer to 16%, with the per-user decline closer to 24%. Still large. Just not 30%.

The other asterisk is geography. This is a worldwide number, and Fishkin himself flags that declines in the US, UK and EU could be partly offset by growth in Asia and Africa. If your audience is mostly American, the global curve is probably too kind to you. Press Gazette's October 2026 ranking found 41 of the top 50 US news websites had fewer visits than a year earlier, with Newsweek down 60% and Fortune down 49%. Different dataset, harsher niche, but it points the same way.

Grade your traffic on a curve (a rough one)

The useful part of this study is that it hands you a market rate to compare against. Not a perfect one. A usable one.

Some rough math, mine, not SparkToro's: a 12.2% decline over about four and a half years is roughly a 3% drop per year in total open-web visits. The per-user decline, annualized over the full window, comes to something like 7% a year. Those are the two reference lines I'd put on any multi-year organic traffic chart.

A micro-example. Say your blog did 100,000 monthly sessions in early 2022 and 84,000 now. That's down 16%, which looks bad on its own. Against a market that's down about 12%, you're trailing by roughly 4 points, which is a much more specific (and fixable) conversation. If you're at 90,000, you're ahead of the open web. Same red arrow, very different meeting.

Think of it like judging a fund manager. Nobody fires a manager for losing 8% in a year the index lost 15%. Traffic reporting should work the same way, and mostly it doesn't.

What I'd actually set up, ideally in under an hour:

  • Add an index line. In whatever dashboard tracks organic sessions, add a reference series that declines about 3% a year from your 2022 baseline. If you're US-heavy or in publishing, make it steeper. I'd use 5% a year as a starting guess for US-only content sites, and I'd hold that number loosely.
  • Report the gap, not the delta. "Organic is 6 points ahead of the market" is a sentence leadership can act on. "Organic is down 9%" mostly invites panic.
  • Track value per session alongside volume. If visits per person are shrinking across the web, the visits you do get are carrying more of the load. Conversions per 1,000 sessions, email signups per 1,000 sessions, pipeline per session. If those are rising while traffic falls, that's a story worth telling, and probably the real one.

One caution on that last point. If your sessions have been padded by bots, cleaning them out makes per-session metrics look better without anything real changing. We covered how direct traffic rose 141% across 12 sites while engaged direct didn't move, and that same distortion can flatter or wreck these ratios depending on which way it runs.

Where the missing visits probably went

Nobody knows for sure, but SparkToro's own earlier work gives a partial answer for search. Its June 2026 zero-click study found 68.01% of US Google searches from January to April 2026 ended without a click, up from 60.45% in 2024. That's browser searches only, so the Google app isn't even in there.

So some share of the decline is search answering questions itself. Some is apps. Some is people simply spending less time browsing than they did when they were stuck indoors. I'd guess the mix differs a lot by category: recipe and how-to sites probably feel the zero-click effect most, B2B software sites probably less.

And to be fair, none of this is brand new. Fishkin points out that 2021 "was hardly the start of the Zero-Click era," with big declines already happening between 2012 and 2020. It just feels less forgiving now, because the slack that used to come from a growing web is gone.

This connects to something we wrote about earlier this month, when data showed email grew 16% while organic fell 13%. If the open web is a shrinking pond, owned channels are the bucket you fill while the water's still there.

A flat line is the new up-and-to-the-right.

A guess for SparkToro's 2027 rerun

Prediction, with stakes: when SparkToro reruns this next year (Fishkin says he plans to recheck annually), I expect top-million visits to land at least 15% below the 2022 peak, and the top 100's share to cross 38%. The 2025 share jump looked like a step change rather than noise, and AI answer surfaces are mostly built on top of the biggest sites. If I'm wrong and traffic climbs back toward 2.1 trillion, it'll most likely be the post-COVID hangover finally finishing, which would honestly be good news for everyone.

For your own reporting, though, the practical change is small. Put the market line on the chart. Stop treating every year-over-year dip as a team failure. And start asking whether each visit is worth more than it used to be, because from what I've seen, that's the question the next budget conversation is going to circle back to anyway.

I don't think anyone gets credit for holding flat in a shrinking market. Maybe they should.