I saw the headline before I saw the data: Windows had apparently lost something like 15 points of market share almost overnight. The reaction online was immediate, screenshots everywhere, Linux fans taking a victory lap. I decided to go find the underlying figure before joining in.
The only problem was that the number was wrong, and the real story was sitting in the data that informed the chart everyone was screenshotting.
StatCounter’s numbers made it look like Windows was in freefall
The screenshot everyone shared
In April 2026, StatCounter had Windows at roughly 79% of the desktop market. By June, that same tracker showed Windows at just 56.55%, a drop of over 22 points in two months. Linux appeared to climb to 4.39% and macOS to 11.89%–16.37% over the same stretch, and the story spread fast, with outlets from Slashdot to Linux fan sites and self-styled AI influencers on X all running with the screenshot.
I get the appeal, and understand the confirmation bias. A 15-to-20 point swing in two months is the kind of number that makes a great post. It also makes zero technical sense. Operating systems don’t lose a fifth of their installed base in eight weeks, and Linux’s modest gain couldn’t possibly account for tens of millions of missing Windows machines. Something else had to be going on.
The missing users were sitting in an “Unknown” column
One line nobody was reading
Right below Windows in StatCounter’s own table sat a category simply labeled “Unknown,” suddenly accounting for 21.45% of desktop traffic. That’s nearly five times Linux’s entire reported share, and it appeared in the same month Windows cratered. Korben’s analysis of the raw numbers pointed straight at it: remove that Unknown bucket, and Windows climbs right back to around 72%.
That’s the trick with StatCounter. It doesn’t count devices or installations. It reads the browser’s user agent string on billions of page views across 1.5 million sites, and anything with a modified, missing, or unrecognized user agent (a bot, a scraper, a privacy tool, or an AI crawler) gets placed into Unknown instead of Windows, macOS, or Linux. StatCounter says it strips out bot traffic and adjusts for pre-rendering, but something got missed in June.
If you want to see this for yourself, don’t trust the chart on the page. Click the download button on any StatCounter breakdown and pull the CSV instead. Tech2Geek found that the exported file recalculates the percentages using only devices it could actually identify, dropping the Unknown category entirely and pushing Windows back to its real range. Neither figure is technically false; they just explain different things about the same traffic.
It also doesn’t mean that StatCounter’s data is worthless going forward. The company caught the error and republished the corrected figures within days, once outlets like PCMag started asking pointed questions about the Unknown category and why it had suddenly ballooned. Month-over-month trend lines built from billions of page views are still genuinely useful for spotting slow shifts in browser or OS usage. It’s this type of single dramatic outlier, the one month that breaks a pattern that’s been stable for years, that deserves a second look before it gets shared.
The trend is real, just not as dramatic
Windows is still slipping, just slowly
Once corrected, Windows sat at 72%, and Windows Latest expected it to drift back toward the high-70s as more data came in. That’s still a decline from a year earlier, roughly five points over twelve months by Tech2Geek‘s count, with Linux and macOS both picking up a small amount of that difference. It’s just nowhere near the collapse the original screenshot from June implied.
This also wasn’t the first time StatCounter made an error. Windows Latest notes that the same service briefly showed Windows 11’s share reversing at the end of 2024, a swing that turned out to be a real but far smaller shift than this June’s numbers suggested. In 2025, StatCounter also reported an outright impossible spike in Windows 7 usage, climbing from under 1% in July to nearly 10% by September, before that figure was corrected back down. Sudden, dramatic swings in any single month of this data are usually a sign something got miscounted, not a sign of doom for any given OS.
Windows also maintains such a dominant position that a real mass exodus would be hard to miss. It commands around 72% of the global desktop market even in a normal month and has an install base built over decades. A shift of that scale would show up gradually across multiple independent trackers, not as a single-month cliff on one dashboard while every other analytics firm reported something closer to normal.
It’s worth remembering, too, that even accurate market share numbers can be misleading on their own. A browser with a tiny single-digit percentage of the market isn’t necessarily struggling, and a huge percentage doesn’t automatically mean an operating system is thriving.
Trust the data, not the screenshots
There’s no reason to write off StatCounter. It’s still one of the largest independent trackers out there, and its long-term trend lines are useful. But it’s still important to investigate dramatic headlines like that one, especially based on a single month of data. It’s smart to pull the raw CSV and check what’s hiding in the categories nobody screenshots. If nothing else, it helps us all become better critical thinkers facing an onslaught of clickable dramatic headlines.