Web traffic decline across the world’s top million websites now stands at 12.2% below the Q1 2022 peak, according to Similarweb data published by SparkToro. Total browser visits came in at 1.970 trillion in Q3 2026, down from a high of 2.243 trillion. The measurement covers desktop and mobile web visits worldwide, excluding app use.
What Does the SparkToro Dataset Actually Show?
SparkToro co-founder and CEO Rand Fishkin requested total visit counts from Similarweb (a web analytics platform that tracks browser sessions via a global panel), broken down by site rank. Total visits peaked at 2.243 trillion in Q1 2022, hit a low of 1.914 trillion in Q1 2026, a 14.7% drop from peak, then recovered slightly to 1.970 trillion by Q3 2026. The trend line holds roughly steady through 2023, dips in early 2024, and falls again in the first half of 2025.
Similarweb divided the top million sites into five ranking bands, from #1-10 to #100,001-1,000,000. The share of visits going to sites outside the top 100 stayed relatively stable across the period. Combined, the top 100 sites held roughly 33% of all visits in Q4 2021 and roughly 36% in Q3 2026. The percentages for each band below the top 100 were slightly lower at the end of the period than at the start. Fishkin described the top-100 share growth as small, writing that “stagnation is a more accurate description of the trend.”
Why the 29.2% Per-User Decline Needs a Caveat
SparkToro also reported a 29.2% drop in visits per internet user, calculated by dividing Similarweb’s total visits by DataReportal’s estimated quarterly internet users globally. That figure needs context before it can be used as a benchmark.
According to DataReportal’s October 2025 report, the global internet user count jumped from 5.65 billion to 6.04 billion in a single quarter (Q3 2025). DataReportal attributed this jump as “largely attributable to significant revisions” in how internet use was officially reported by India and China, and explicitly warned readers not to interpret the change as hundreds of millions of new people coming online in 90 days.
SparkToro’s post did not include this context. Their estimate of 24% user growth compared figures from before and after that revision. The more stable headline figure remains the 12.2% decline in total visits, which does not depend on the user-count series at all.
What Could Be Driving the Web Traffic Decline?
Fishkin did not offer a single cause. His list of possible factors includes:
- Increased app-only use (Similarweb’s panel tracks browsers, not apps)
- Reliance on search engine results pages without clicking through to sites
- AI tool utilization
- Social media platforms replacing direct site visits
- Behavior changes following the COVID-19 pandemic
Fishkin added that Big Tech consolidation is likely much more pronounced in apps, a channel entirely absent from this data. He also noted that because the data is worldwide, it is plausible that traffic to U.S., U.K. and EU sites declined while regions with growing internet access, such as Asia and Africa, grew. The dataset does not break out regions, and the post does not state the size of Similarweb’s panel.
A Note on the Corrected Figures
SparkToro updated their post after initial publication. On Nov. 10, Fishkin said on X that the prior version “had an incorrect graph/data.” The original headline reported a 24% decline. The corrected figure is 12.2%. The post’s URL still reflects the original 24% number. SparkToro sells audience research software, and Fishkin co-authored the book “Zero Click Marketing,” Search Engine Journal noted.
What This Means for AI-Search Visibility
A drop in browser visits does not automatically mean a drop in brand exposure. That distinction matters more now than it did in 2022.
Similarweb’s panel counts visits to websites. If someone asks an AI assistant for the best accounting software for a small business and gets an answer naming three products, the assistant’s own site may log a visit, but the three brands do not, unless the user clicks through. The brands were still seen; their sites just do not show it.
“AI tool utilization” appears on Fishkin’s list of possible explanations for the drop. He does not claim it caused the decline. In Hingewise’s view, however, the structural gap in measurement is worth naming directly: browser-visit metrics were designed for a web where content discovery ends in a click. That is no longer the only path.
The practical implication: a brand’s own site could lose visits while the brand appears more often in AI-generated answers. These two outcomes are not contradictory. They reflect a split in discovery channels, one that browser-visit data captures and one that it does not.
Before using this data to benchmark your site’s traffic, consider checking:
- Whether your analytics window matches the reference period (Q1 2022 through Q3 2026)
- Whether your primary audience is in a region, such as Asia or Africa, where internet access grew significantly during this period
- Whether your traffic figures include app-based visits (Similarweb’s panel excludes apps)
- Your site’s approximate rank among the top million, to identify which ranking band applies to your traffic share
- Whether your zero-click and AI referral data point in the same direction as your browser-visit trend
The 12.2% figure is accurate for what it measures. It is best read as global context for a multi-year trend, not as a direct benchmark for any single market or site. Fishkin plans to compare the findings against other sources and revisit the data every year or two. Regional breakdowns, if they become available, will make the figures considerably more useful.
Sources: SparkToro, Rand Fishkin, Nov. 10, 2026; Search Engine Journal, Nov. 12, 2026
Lam Nguyen · Hingewise
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.
