AI crawler IP ranges

Every major AI company publishes the IP ranges its crawlers use. This page reads those files once a day and shows what is in them, with a link to each source so you can check it yourself.

Not sure which crawler is which? See what each one is for.

Last read: August 24, 2026 · 2,448 ranges tracked

Google

1,641 ranges

Googlebot is search. Google-UserTriggered covers fetches a person asked for. Google-Extended is the one that governs Gemini training, and it is published separately.

Crawler	Ranges	IPv4	IPv6	Last read	Source
Googlebot	315	169	146	August 24, 2026	Published file
Google-SpecialCrawlers	270	135	135	August 24, 2026	Published file
Google-UserTriggered	1056	528	528	August 24, 2026	Published file

DuckDuckGo

481 ranges

DuckAssistBot supports DuckDuckGo's assistant answers. The published list is large because it covers shared infrastructure.

Crawler	Ranges	IPv4	IPv6	Last read	Source
DuckAssistBot	481	481	0	August 24, 2026	Published file

OpenAI

260 ranges

GPTBot collects training data. ChatGPT-User and OAI-SearchBot fetch pages when someone asks ChatGPT about them, so blocking those two removes you from answers.

Crawler	Ranges	IPv4	IPv6	Last read	Source
ChatGPT-User	204	204	0	August 24, 2026	Published file
GPTBot	21	21	0	August 24, 2026	Published file
OAI-SearchBot	35	35	0	August 24, 2026	Published file

Microsoft

28 ranges

Bingbot serves Bing search and Copilot answers.

Crawler	Ranges	IPv4	IPv6	Last read	Source
Bingbot	28	28	0	August 24, 2026	Published file

Anthropic

26 ranges

ClaudeBot collects training data. Anthropic began publishing a machine readable feed in August 2026, before which verification relied on a documentation page.

Crawler	Ranges	IPv4	IPv6	Last read	Source
ClaudeBot	26	26	0	August 24, 2026	Published file

Perplexity

12 ranges

PerplexityBot indexes pages. Perplexity-User fetches a page because someone asked about it in a conversation.

Crawler	Ranges	IPv4	IPv6	Last read	Source
PerplexityBot	8	8	0	August 24, 2026	Published file
Perplexity-User	4	4	0	August 24, 2026	Published file

Telling a real crawler apart

Why is the user agent not enough?

A user agent is a text field the client chooses. Anyone can send one that says GPTBot. We regularly see traffic claiming to be a major AI crawler arriving from hosting providers those companies do not use. Checking the IP against the vendor's published list is what separates the two.

How do I verify a request myself?

Take the client IP and check whether it falls inside one of the ranges in the vendor's own file, linked in the table above. If it does not, the request is not from that crawler no matter what the user agent says. Some vendors also support a reverse DNS lookup that resolves to their domain, followed by a forward lookup back to the same IP.

Can I rely on my CDN's verified bot flag?

Only if you know it is populated. On Cloudflare, the verified bot field is present on every plan but is only filled in on plans with the Bot Management add on. On the others it is present and always false, so treating false as proof of a fake will condemn genuine crawler traffic. Check what your plan actually provides before trusting the field.

How often do these ranges change?

It varies by vendor and it is not steady. Most lists sit unchanged for weeks. OpenAI moves the most: in the first week of August 2026 it added 47 ranges and withdrew 62 across its three crawlers. A list you cached last month may already be wrong.

Where does this page get its data?

Directly from the files each vendor publishes, linked in the table. We read them once a day and record what changed. We do not edit the contents, and nothing here comes from any customer's traffic.

Ranges are read from the vendor files linked above. Those files are the authority. If one of them disagrees with this page, trust the vendor.