Will a wildcard robots block stop Google-Safety?

No. Google says this agent is unaffected by crawling preferences, which includes a `User-agent: *` disallow.

Is Google-Safety part of normal Search indexing?

No. Google-Safety handles abuse checks. Googlebot is the crawler associated with ordinary Search discovery and has different robots behavior.

How can I reject a genuine Google-Safety fetch?

Robots.txt will not do it. Require authorization or enforce a denial in the CDN, firewall, or application, knowing that Google will be unable to complete that safety check.

Does Google-Safety collect AI training data?

No training role is documented. Its stated task is abuse-specific fetching, not collecting pages for AI search or a model corpus.

Google-Safety: What It Is & How to Handle It

What is Google-Safety?

A Google-Safety request is an abuse check. Google's own example is malware detection for a link posted publicly on a Google property. That makes the visit different from Googlebot's routine work of discovering pages for Search.

Google identifies the client with Google-Safety inside a browser-like header. It belongs to Google's special-case crawler class, whose requests can be checked against the published special-crawler address ranges. Google's documented reverse and forward DNS procedure offers another way to confirm the source.

Crawl preferences do not govern this agent. Google states that Google-Safety is unaffected by robots.txt, so neither a wildcard disallow nor a named group is expected to prevent the fetch. A server can still require authentication or reject the request at the network edge.

The documented purpose is safety investigation, not AI discovery. Google does not say that Google-Safety retrieves material for generated answers or model training, and the agent is not evidence that a URL entered the regular Search index. Blocking it only prevents that safety request from reaching the resource.

Not relevant for AI search

Is Google-Safety relevant for AI search?

No. Google-Safety is not part of AI search or training, so allowing or blocking it does not change your AI visibility.

Google-Safety runs synthetic monitoring or uptime checks on a schedule. It is not collecting content for any index or model, so it has no bearing on search rankings or AI answers. You usually see it because your own team or a service you use set up a check.

How to handle Google-Safety

Treat an isolated visit as a possible abuse-related fetch, then verify it with Google's special-crawler network data or forward-confirmed reverse DNS. Review the requested path for a publicly shared download or another resource that could have prompted a safety check.

This group can label the intended policy, but Google says it will not control the agent:

User-agent: Google-Safety
Disallow: /

Use real authorization, an edge block, or an application response if access must be denied. Such a block also prevents Google's safety system from examining that URL, so it is not a useful way to manage ordinary crawl volume.

Examples

Someone posts a download link on a Google property. The destination receives a Google-Safety fetch even though the site's wildcard robots group disallows all crawlers.
An access log shows one Google-Safety request for an executable file. The analyst verifies the address against Google's special-crawler data before investigating the file.

Frequently asked questions about Google-Safety

Learn about AI visibility monitoring and how Promptwatch helps your brand succeed in AI search.

Google uses the agent for abuse-specific fetching. The example in its documentation is malware discovery for links posted publicly on Google properties.