adtestbench Higgsfield
EN

AI news · Google · Google-Agent

Google tells sites how to let its AI agent past their bot filters

A new help page explains how firewalls and bot checks can recognise Google-Agent, which browses, fills baskets and books on a person’s behalf.

By Alexandre S. , 07:30 UTC

On Tuesday 17 November 2026 Google posted instructions for site owners whose security tools keep out Google-Agent, the name its AI agents use when they browse and act for a person. The page covers bot-detection services, web application firewalls (filters that screen incoming traffic) and the content delivery networks that serve a site’s pages. Google’s stated goal is that a person’s genuine requests get through without a site lowering its defences, Search Engine Roundtable reports. Google-Agent itself has been documented since March; the guide is new.

Retro-futurist illustration: a striped sunset over a grid horizon under a starry sky, with a software window standing on the horizon.
Drawn by adtestbench from “Google lists six ways sites still block its AI agent after allowing it”,

PPC Land’s reading of the page explains why shops and travel sites should care. The agent makes bookings, fills baskets and searches and filters a site’s listings, by Google’s own description, and it generally ignores robots.txt, the file sites use to steer crawlers. A site confirms the visitor is Google’s agent in up to three ways: the name it gives in its request (the user-agent header), whether its address appears on Google’s published list, and, for some requests in an experiment Google is running, a digital signature. Google tells sites to let traffic that passes those checks skip bot challenges such as CAPTCHA puzzles.

The page also lists six ways a site can keep refusing the agent after adding an allow rule. One example: Google sends many people’s requests out through a shared set of addresses, so a cap on requests per address can shut the agent out.

Google’s guide expects its agent to add items to baskets and make bookings, so a retailer’s bot defences now decide whether those agent-led visits reach checkout, since the agent generally ignores robots.txt.