Google-Extended has to be allowed to be eligible for grounding
Blocking both training and grounding narrows the routes by which Gemini can use a document as evidence. Navirang's own site allows both, because citation is the goal.
GEMINI
Gemini citation reference —
a different control point
from AI Overviews
Same company, different control point. The Google-Extended token applies to Gemini; AI Overviews visibility is decided by Googlebot.
SHORT ANSWER
Gemini is Google's conversational answer service. It is a different product from the AI Overviews inside Google search results, and more importantly it has a different control point — the Google-Extended token in robots.txt governs training and grounding for the Gemini app and Vertex AI. Blocking that token restricts use on the Gemini side but has no effect on Google search or AI Overviews.
Google-Extended is the control pointA separate product from AI OverviewsNo effect on the search index
SPEC
Gemini specifications
Only items that can be verified are listed. Rows with a verification method on the right can be reproduced yourself.
- Operator
- Control token
- Google-Extended. Not a crawler but a robots.txt-only control token, with no User-Agent of its own You will never find a visit under this name in your server log — this one is verified in robots.txt, not in logs.
- What it governs
- Training and grounding for the Gemini app and Vertex AI, and training of the generative AI response models in search
- What it does not govern
- The Google search index and ranking, and AI Overviews visibility. Those are decided by Googlebot
- Actual crawling
- Googlebot does the collecting. Block Googlebot and you disappear from the entire Google ecosystem
- Relation to AI Overviews
- A different product with a different control point. Visibility in the two has to be measured separately
- Korean
- Supported
CONDITIONS
What it takes to be cited in Gemini
These are the items that differ most in this engine. Principles common to every engine are collected on the answer engines hub.
Googlebot has to be able to read it first
Collection itself is Googlebot's job. If that side is blocked, there is no raw material regardless of the Google-Extended setting.
It has to be measured separately from AI Overviews
Even within Google the two products return different results. Inferring one from the other leads the judgement astray.
MYTHS
Common misconceptions
Only the ones we meet repeatedly in audits.
-
“Blocking Google-Extended lowers your Google search ranking”
It does not. There is no effect on the search index or ranking; only use on the Gemini side is restricted. Blocking Googlebot, by contrast, removes you from search itself. Navirang corrected its own comment on this distinction on 2026-07-31.
-
“If you appear in AI Overviews you appear in Gemini”
They are separate products with different control points and different results. In our measurements each counts as its own cell.
-
“You can find Google-Extended visits in your log”
You cannot. It is a control token with no User-Agent of its own, so it never appears in logs. This item is verified in robots.txt only.
HOW TO CHECK
How to check it yourself
A Navirang audit follows the same order. There is nothing stopping you from running it internally first.
- 01
Check the Google-Extended group setting in robots.txt (it will not appear in logs).
- 02
Check that Googlebot is not blocked while you are there — collection happens on that side.
- 03
In a logged-out session, ask the target question in Gemini, then ask the same question in Google search and compare against the AI Overview result.
- 04
Record the two results on separate lines of the same table.
FAQ
Questions about Gemini
Each answer is written to be quoted as it stands.
Are Gemini and Google AI Overviews the same thing?
No. AI Overviews is a generated summary attached to Google search results; Gemini is a separate conversational service. Above all the control points differ — Google-Extended in robots.txt applies to training and grounding for the Gemini app and Vertex AI, while AI Overviews visibility is decided on the Googlebot side. In Navirang's measurements each counts as its own cell and the results are recorded separately.
Should we allow Google-Extended?
If citation is the goal, we recommend allowing it. The token decides whether the Gemini app and Vertex AI may use your documents for training and grounding. Blocking it leaves Google search visibility unchanged, so it can be chosen without worrying about search loss. That said, a publisher whose content is itself the product may not want to be used for training, so the judgement varies with the business model.
Google-Extended does not appear in our server log.
That is normal. Google-Extended is not a crawler but a token interpreted only in robots.txt, so it has no User-Agent of its own and leaves no visit record in access logs. The one actually fetching documents is Googlebot. This item is verified through the robots.txt setting, not through logs.
RELATED
Related reading
The canonical crawler list is kept as a table in a separate article — this page carries only the rows for this engine.
Last verified August 27, 2026
How does your brand look in Gemini right now?
Send us a URL and our free audit asks the questions in all 7 engines and reports what actually came back, recorded engine by engine.
We reply within one business day.