Google-Agent is the name Google's agents use to identify themselves when they visit a site and take actions because a person asked them to. It is not the robot that indexes pages for Search: it is an assistant working on someone's behalf, so it can search, filter, open product pages and reach the cart.
On November 17, 2026 Google published a technical guide for sites to let Google-Agent through. Here is what it asks for, what it leaves out and what to check in your store before your security provider decides for you.
What Google-Agent is and how to recognize it
Google describes it as the agent its services use to browse the web and take actions when a user asks. The guide does not say which product uses it or which actions it can complete, so it is best not to assume.
A Google-Agent visit can be recognized by three signals:
- Its name. Every visit says "Google-Agent" in its identification. Anyone can copy that name, so on its own it proves nothing.
- Its origin. It comes from Google addresses published in an official list. A visit that claims to be Google-Agent but comes from another address is not.
- Its signature. Google is testing a digital signature on each visit that the site can verify (the standard is called Web Bot Auth). It is experimental: not every visit carries it yet.
Why it matters if you sell online
Behind every Google-Agent visit there is a person who asked it for something: find a product, compare prices, check stock. If your site shuts the door, the agent comes back with another store's answer.
The risk is not blocking on purpose. It is blocking without knowing. In our study Agentic commerce in Mexico 2026, 9 of 12 sites have no explicit agent policy: whoever decides is the firewall default. And in the sample, two sites stopped the agent that identifies itself while letting the anonymous one through.
That is exactly the case Google's guide wants to avoid: punishing the visitor who says who it is.
What Google asks for
- Check with your security provider first. If your CDN, firewall or bot protection service already recognizes Google-Agent, follow its instructions. If not, your team can verify it with the three signals above.
- Skip the challenges, not just the blocks. Letting it through includes skipping CAPTCHAs and automated checks, including the ones that appear on search or form submission.
- Open the whole journey. Search, filters, product pages, cart and booking, plus the internal services those pages call behind the scenes. If the agent reaches the home page but cannot filter, it does not get far.
- Put the exception before broad blocks. It has to apply before blocks on cloud networks, managed bot rules and category blocks on "AI agents".
- Do not rate limit it by address. Many people use Google-Agent at the same time from the same Google addresses. A limit designed for one person throttles all of them.
- Keep intermediaries from erasing its identity. If your site sits behind load balancers or proxies, they must preserve the origin address, the identification and the signature.
What the guide leaves out
- Whether it respects robots.txt. The guide does not mention it. On another page, Google explains that because the visit was requested by a user, these programs generally ignore
robots.txtrules. In practice,robots.txtis not the tool to decide whether Google-Agent gets in: that decision lives in the firewall. - Whether it affects your rankings. The guide says nothing about rankings or indexing. Google-Agent is not the Search robot, and nothing in the documentation suggests that letting it through moves you up or down in Google.
- Which product uses it or whether it pays. It names no products and does not say whether the agent completes payments. There is no basis today to claim it already buys.
What to check this week
- Ask your security provider whether it already verifies Google-Agent and what it does with those visits today.
- Check whether category blocks such as "AI agents" or "automated bots" are on by default.
- Walk the full path with your technical team: search, filters, product page and cart. Every step with a CAPTCHA is a step where the agent stops.
- Review per address rate limits and exempt verified Google addresses.
- Put the policy in writing. Letting Google-Agent through is a business decision: leadership should make it, not the default settings.
The Edgebound ACR scanner checks, among other things, whether your store blocks the agent that says who it is and lets the anonymous one through.
Frequently asked questions
What is Google-Agent?
It is the name Google's agents use to identify themselves when they visit a site and take actions because a person asked them to. It can be recognized by its name, by coming from addresses Google publishes and, experimentally, by a verifiable digital signature.
Is Google-Agent the same as Googlebot?
No. Googlebot crawls the web for Search. Google-Agent visits a site because a person asked it for something, like finding or comparing products. They are different visits and identify themselves separately.
Does Google-Agent respect robots.txt?
The guide does not say. In its documentation, Google explains that visits requested by a user generally ignore robots.txt. To decide whether it gets in, the tool is the firewall or the security provider.
Does letting Google-Agent in improve my Google rankings?
The guide says nothing about rankings or indexing. What does change is whether the agent can finish on your site what the person asked for, or has to go somewhere else.
How do I know a visit really comes from Google-Agent?
With three checks: it identifies itself as Google-Agent, it comes from an address in Google's official list and, if it carries a signature, the signature is valid. The name alone is not enough, because anyone can copy it.
Limits
- This note summarizes Google's public documentation as of November 17, 2026. The guide is new and the digital signature is experimental: it may change.
- For this note we did not test Google-Agent against real stores.
- The figures on Mexican sites come from a single run of our study, with a small sample that is not representative.
Disclosure of interest
Edgebound sells digital commerce architecture and integration services, and the ACR scanner, which can address what this note describes. This note was not paid for or reviewed by Google.
Sources
- Google Search Central. Allow Google-Agent traffic. developers.google.com
- Google Search Central. Google's user-triggered fetchers. developers.google.com
- Google Search Central. Web Bot Auth. developers.google.com
- Google Search Central. Crawling documentation changelog, November 17, 2026. developers.google.com
- Edgebound Labs. Agentic commerce in Mexico 2026.
What does an AI agent find when it enters your store?
The ACR scanner measures it with evidence your team can reproduce.
See the scanner