AI TrainerGoogle
Last verified: 2026-07-16 · maintained by Unsourced
Google-CloudVertexBot fetches pages on behalf of Vertex AI customers building Gemini-grounded applications. It identifies from Google's published ranges.
Google-CloudVertexBot is a fetch made on behalf of a Vertex AI customer: when a business builds a Gemini-grounded app or agent that needs to read a specific page, this is what retrieves it. Its requests reflect a third party's application logic rather than Google's own indexing, and they arrive from Google's published IP ranges, so the bot is straightforward to verify. It is distinct from Google-Extended — one controls AI-training permission, this one actually fetches pages for live Vertex-powered applications.
Google doesn't expose a reverse-DNS hostname for Google-CloudVertexBot, so verification rests on its 270 published IP ranges: the request is genuine only when its source address sits inside that set. A Google-CloudVertexBot user-agent arriving from any other IP is spoofed — on its own the header proves nothing.
Recommended: keep. Unlike the Google-Extended training switch, this one fetches pages live for a third party's Vertex AI app, so the real question is whether you want Gemini-grounded business apps to read your content.
If you do choose to act in robots.txt (which crawlers honour but don't enforce):
# Google-CloudVertexBot: recommended to ALLOW — blocking can cost you AI visibility User-agent: Google-CloudVertexBot Disallow:
What triggers Google-CloudVertexBot?
A Vertex AI customer's application. When a business builds a Gemini-grounded app that needs to read a specific page, this is the fetch that retrieves it — so its requests reflect a third party's app logic, not Google's own indexing.
Is Google-CloudVertexBot the same as Google-Extended?
No. Google-Extended is a training-permission token; Google-CloudVertexBot actually fetches pages live for customers' Vertex-powered apps. One controls AI-training use, the other is real-time retrieval.
How do I verify Google-CloudVertexBot?
It arrives from Google's published IP ranges, so match the source IP against those — a request from outside Google's allocation isn't genuine.
Unsourced checks every crawler against operator-published ranges and forward-confirmed reverse DNS, and shows which AI assistants cite you.
Check your site free →10-day free trial · no card required · cancel anytime