New: monitor which AI crawlers actually visit your site with Sona Agent Analytics  |  See the platform →

Google-Agent

Google's agentic fetcher - Gemini navigating and acting on a user's behalf. Documented to ignore robots.txt.

Reference reviewed by Sona

OperatorGoogle
PowersGemini agentic browsing and task execution
PurposeUser-triggered fetch
User-agent tokenGoogle-Agent
Respects robots.txtNo (user-triggered)

Google-Agent identifies fetches made when an AI agent on Google infrastructure navigates the web and performs actions a user requested. Google's own phrasing - navigating and performing user-requested actions - is what separates it from the rest of this directory: this is not a bot reading pages for an index, it is software using your site.

That distinction has consequences beyond content policy. An agent following a multi-step task can move through navigation, forms, and interactive flows rather than fetching one URL and leaving. Rate limits, bot-detection heuristics, and checkout or signup funnels built on the assumption of a human at a keyboard are all touched by this traffic in ways a normal crawler never reaches.

Google documents it among the user-triggered fetchers, which "generally ignore robots.txt rules" because a person initiated the request. Server- or firewall-level user-agent filtering is the only mechanism that actually enforces a block.

Commercially, agentic traffic is the category most likely to carry intent - an agent completing a task on someone's behalf is closer to a customer than to a crawler. Blocking it protects you from automated interaction you did not design for; it also blocks a would-be buyer whose agent was sent to your site.

How Google-Agent behaves

  • Navigates and performs actions rather than only reading - can traverse forms and multi-step flows.
  • Documented among Google's user-triggered fetchers, which generally ignore robots.txt.
  • Sends mobile and desktop user-agent variants carrying the Google-Agent identifier.
  • May generate several sequential requests per task, unlike single-page fetchers.

Allow Google-Agent

Gemini users whose tasks touch your site get real content and working interactions - and agentic visits often carry direct commercial intent.

User-agent: Google-Agent
Allow: /

Block Google-Agent

You want no automated agents interacting with your site - enforce it in your WAF, since robots.txt is documented as generally ignored.

User-agent: Google-Agent
Disallow: /

Heads up: Google-Agent does not reliably honor robots.txt. To enforce a block, combine the rule above with server- or firewall-level filtering of the user-agent.

How to verify Google-Agent

Published IP range feed

Verify against Google's user-triggered-fetchers IP feed, which is separate from the Googlebot and common-crawlers files. Verification matters more for agentic traffic than for ordinary crawls: because this token can drive forms and multi-step flows rather than just read pages, treating an unverified request as a legitimate Google agent means granting interactive access on the strength of a header anyone can set.

Documented reverse-DNS suffixes: .google.com

Check an IP against this bot

Commonly confused with Google-Agent

GoogleAgent-URLContext

Similar naming, different job: URL-context fetches specific URLs a developer supplies to the Gemini API, while Google-Agent navigates to complete a task.

Gemini-Deep-Research

Deep Research gathers sources for a report; Google-Agent takes actions. Both are Gemini-driven and both ignore robots.txt.

Meta-ExternalFetcher

Meta's agentic fetcher, with the same weak robots.txt guarantee and the same interactive-traffic implications.

Google-Agent FAQs

How is Google-Agent different from a crawler?

Crawlers read pages for an index. Google-Agent navigates and performs actions a user asked for, so it can move through forms and multi-step flows rather than fetching a single URL - which makes it relevant to your bot detection and funnels, not just your content policy.

Can robots.txt block Google-Agent?

Not reliably. Google documents its user-triggered fetchers as generally ignoring robots.txt because a person initiated the request. Enforcement requires user-agent filtering at the server, CDN, or WAF level.

Should I be worried about agents hitting my forms?

It is worth testing rather than worrying. Rate limits, bot-detection rules, and signup or checkout funnels built for human timing can behave unexpectedly under agentic traffic - a good reason to verify how your flows respond.

Is blocking agentic traffic a good idea commercially?

Consider what you are turning away. An agent completing a task for someone is closer to a prospective customer than to a crawler, so a block protects you from unplanned automation at the cost of a potential buyer's proxy.

Can Google-Agent read your page right now?

Test any URL and see exactly what AI crawlers receive.

Check my site