Official crawler IP ranges: 18 lists and an IP checker

The 18 JSON lists

Every file has the same shape: a prefixes array whose items hold an ipv4Prefix or an ipv6Prefix.

The date is the timestamp the operator put inside the file. Bing's file carries 2024-01-03 and still holds 29 ranges on 2026-11-08, so read the date as "last stamped by the operator", not as proof that a list is current.

Amazon: three pages, not JSON files

Amazon lists its addresses on three pages of its developer site. Each page embeds the data as JSON text, and the pages use different field names (ipv4Prefix for single addresses, ip_prefix for CIDR blocks), so the script does not read them. The counts are the number of entries found in the page text on 2026-11-08.

Who publishes no list

• Meta: the crawler page gives user agent strings for meta-externalagent, meta-externalfetcher, Meta-WebIndexer and facebookexternalhit but no list of IP ranges (checked 2026-11-08). Requests that call themselves Meta can only be judged by their user agent.
• ByteDance (Bytespider) and MistralAI-Training: no list documented on the date above.
• Google-Extended and Applebot-Extended are robots.txt control tokens. They do not crawl, so they have no IP ranges.

Check one IP

python check_ip.py 66.249.66.1 2001:4860:4801:10::1 132.196.86.1 8.8.8.8

The script downloads the 18 lists listed in the CSV and prints the matching list for each address. Output from the run on 2026-11-08:

66.249.66.1                              google-common (Googlebot and other common Google crawlers)
2001:4860:4801:10::1                     google-common (Googlebot and other common Google crawlers)
132.196.86.1                             openai-gptbot (GPTBot)
8.8.8.8                                  no match in 18 lists

Run it on the IPs behind a suspicious user agent in your own log. An address that calls itself Googlebot and matches nothing is not Google. In one 30-day audit of a single site, 531 of 2,278 requests that called themselves Google came from 21 IP addresses outside Google's published ranges.

Keep it fresh

Operators change these files, and some addresses move. The CSV is a snapshot of 2026-11-08. check_ip.py downloads the lists on every run, so only the counts and dates in the CSV age. The log audit kit does the same download for 13 of these lists before it classifies a log.

Sources