March 29, 2026
Botpocalypse Now
The bot situation on the internet is worse than you could imagine
Half the web is bots? Commenters shout ‘70%’, fret privacy, and hail a simple puzzle that stopped them
TLDR: An art site’s bait pages drew millions of scrapes, and a small proof‑of‑work “puzzle” choked them to almost nothing. Commenters cheer the hard data, argue over the ethics of sharing logs, roast ironic link‑spammers, and debate whether bots are half the web—or closer to 70%.
Glade Art claims it lured bots into digital “tar pits” and watched them face‑plant: over 6.8 million hits poured into bait pages, then a tiny proof‑of‑work “puzzle” (think a quick computer task) called Anubis slammed the door. Traffic plummeted from hundreds of thousands a day to… about 11. Cue the comment section going full Botpocalypse.
One camp is unimpressed and uneasy. “Nothing new,” sighs one commenter, while another flags a big yikes: the site shared a massive request log, which some call a privacy faux pas if any legit humans got swept in. Then the meta‑drama: users spotted a brand‑new account shamelessly link‑spamming its app in the very thread about bots. The irony writes itself, and the snark peaked with a single‑line mood setter: “The final Eternal September.”
On the other side, data fans are fist‑pumping. “Finally, experiments, not vibes,” says one, praising how PoW flattened bot traffic. Another vows to add Anubis site‑wide and even float an “llms.txt” file so “good” AI crawlers can read while shady scrapers get iced. The big fight: is the internet 51% bots, or 70%+ once you count scrapers sneaking in from residential connections and skipping site scripts? Jokes fly about feeding bots junk, CAPTCHAs’ jacked cousin, and a web where the machines aren’t just learning—they’re refreshing your page more than your mom.
Key Points
- •Glade Art logged 6.8 million requests to a honeypot endpoint over 55 days, showing heavy scraper activity.
- •Bots observed often ignore robots.txt; the site disallows tar-pit pages to avoid affecting compliant crawlers.
- •Most bot origin IPs were residential/mobile, largely from Asian and Indonesian countries, not datacenters/VPNs.
- •Scrapers typically did not execute JavaScript when crawling random sites, making PoW challenges highly effective.
- •Enabling the Anubis PoW gate at lowest difficulty reduced requests from hundreds of thousands per day to about 11 in 24 hours.