What we ran
- Build: a script downloads each owner’s published list: Google’s crawler files, Bing, OpenAI, Perplexity, Apple, Amazon, AWS, Google Cloud, Azure, Oracle, Cloudflare, Fastly, DigitalOcean, Linode, the Tor exit list, and two public repositories for other datacenter, VPN and proxy ranges. The source table is on the bot traffic checker page.
- Flatten: overlaps are removed, one label per address, the most specific name wins (crawler before cloud before generic hosting).
- Where: our test box, plain fetch, a few seconds. The result ships as one 113 KB file the checker reads in the browser.
- Test on real names: we looked up the address of
www.for the 100 online stores of anti bot software market share.
What we found
| kind | ranges | IPv4 addresses covered |
|---|---|---|
| datacenter or hosting (public lists) | 43,608 | 221,344,473 |
| cloud (AWS, Azure, Oracle, Google Cloud and more) | 7,504 | 197,770,897 |
| VPN | 6,869 | 3,303,811 |
| crawler (named bots) | 2,085 | 130,142 |
| Tor exit nodes | 632 | 1,299 |
| proxies bots used in the last 5 days | 281 | 109,056 |
| all | 60,979 | 422,659,678 |
AWS alone is 4,143 ranges, Azure 2,353, Oracle Cloud 485, Google Cloud 208.
9.8% of the IPv4 space sits in a datacenter, cloud, VPN, Tor or crawler range. IPv6 ranges are matched on their first 64 bits.
On the stores: 66 of 100 shop sites resolve into a named range: 30 generic hosting, 19 Cloudflare, 9 Fastly, 8 AWS. The other 34 sit on networks not on any list.
How you’d use it
- Ad network or publisher: label each click or hit: crawler, datacenter, VPN, Tor, none. Datacenter hits on a campaign are the first thing to ask a buyer about, see invalid traffic.
- Analytics team: paste the server log into the bot traffic checker and read the share of datacenter hits next to GA4 bot signs.
- Data team: the same build, for the sources you care about, refreshed daily with a diff, is a custom feed.
It cannot see residential proxies or botnets on home lines. They look like visitors.