Cloudflare says bots beat humans online — its own pages disagree by 30 points

Cloudflare Radar's traffic view shows automated requests edging out humans. Then a reader opened the Bots page and found 65.9% human, 34.1% bot. Same company, same week, opposite answers.

A dashboard that says the bots have won

Cloudflare Radar generic preview card
Cloudflare Radar generic preview card

Cloudflare Radar's traffic view reports that automated requests now rival or exceed human ones — the latest data point in the long "dead internet" anxiety. The crucial fine print sits in the methodology line itself: the figure is the "percentage of HTTP requests classified as bot (automated) or human," filtered to HTML responses. So it's not bot-vs-human traffic in the wild; it's bot-vs-human among page loads, as judged by Cloudflare's own classifier. The source is a live Cloudflare Radar panel rather than a written argument, which is exactly why the thread carried the analysis.

Two Cloudflare pages, a 30-point gap

The sharpest catch is internal contradiction: Cloudflare's own properties disagree.

"On the Traffic page it is showing Bots more than Human, but on the Bot page it's the opposite: 65.9% Human vs 34.1% Bot" — ChrisArchitect, a web dev and ops engineer

The "for the first time" framing also doesn't survive its own timeline:

"The graph seems like it only goes back to April 27 and on that day it was 57% bot…" — a reader reading the axis

And the headline metric is, several argue, structurally meaningless because the classifier folds blocked humans into the bot bucket:

"there is a difference between 'bot vs human traffic' and 'traffic that CF thinks is bot/human'. Every time CF blocks me, I assume it claims I'm a bot in this chart." — a reader on the classifier's blind spot

The per-country numbers drew the loudest "wait, what": Gibraltar at 92% bot traffic, Iran at 76.9%, Singapore 76.4%.

"Lol, what is happening?" — BugsJustFindMe, reading the country leaderboard

For outside corroboration, one reader brings a non-Cloudflare baseline that actually supports the broad trend even if the dashboard is sloppy:

"According to the Thales Bad Bot Report, in 2025 >53% of traffic came from bots. 2024 was 50-50, and in 2013, it was measured at 43%." — a reader citing the Thales report

A concrete mechanism for why surfaces too: certificate transparency logs make fresh hosts instantly discoverable. "The second I added an SSL cert it went from 0 traffic to 1000 requests/min," reported one operator.

The trend is real; the scoreboard isn't the proof

Bots overtaking humans is plausible and the Thales numbers back the arc. But the specific dashboard going around proves less than its headline: it counts only HTML, only what one vendor's classifier labels, and disagrees with that same vendor's other page by 30 points. The dead-internet vibe is doing the persuading here, not the measurement — and a metric that files every blocked human under "bot" was always going to find a lot of bots.

Get breakdowns like this every day

The top Hacker News stories, unpacked — by email, Telegram, or RSS.

Next story · 7 of 19 Read more
🌍Other

Patreon tells a court order to "stuff it" over a $200k Lego feud

A YouTuber's investigation into a missing $200,000 Star Wars Lego collection escalated into a no-bail warrant, a flight to Mexico, and a lawsuit — and then Patreon's CEO publicly refused to take the videos down. The thread's one real contribution: a close read of the court order explaining why Patreon may be right.

★ 49 · 💬 4 · kotaku.com · 3 min