Gemini

What does Gemini fetch from a website?

Nothing you can see under its own name. Of the 143,601 AI-crawler requests attensira.com received between 5 September and 4 October 2026, read over the Attensira MCP by the user agent each request claimed, none came from an agent filed under Google Gemini. Every Google request in the window came in as Googlebot; those counts are on AI Overview, which is fed by the same crawler.

The zero is not neglect. It is how Google built Gemini. Google's crawler documentation says Google-Extended, the control for Gemini, has no separate user agent string: crawling is done with Google's existing agents. When Gemini searches the web, it draws on Google Search's index, which Googlebot fills.

So two habits from other AI engines do not transfer. You cannot read Gemini's interest in you from your logs, and you cannot block "the Gemini crawler" by name. Our own workspace reads ChatGPT and Perplexity answers, 4,524 of them in that window, not Gemini's, so we quote no Gemini answer rate here.

What is Gemini?

Two things share the name:

• The models. Google's family of multimodal AI models, handling text, images, audio and video. Google states many Gemini models accept context windows of 1 million or more tokens. They power the Gemini app and AI features across Google, including AI Overviews and AI Mode.
• The app. Google's AI assistant, where people ask questions and get written answers, often grounded in Google Search.

How does Gemini use the web?

Through grounding. Google's Gemini API documentation describes grounding with Google Search as connecting the model to real-time web content, so it can cite verifiable sources beyond its knowledge cutoff. Without grounding, Gemini answers from what the model learned in training.

That makes Google's index the gate. A page that Googlebot has not crawled or Google has not indexed cannot be found by a Gemini search step. See crawling and indexing.

What does Google-Extended control?

Google-Extended is a token you can name in robots.txt. Google states it governs whether content is used:

• to train future generations of Gemini models;
• for grounding, meaning content from the Google Search index supplied to the model at answer time, in Gemini Apps and Vertex AI.

And it states what Google-Extended does not do: it does not affect a site's inclusion in Google Search, and it is not a ranking signal. So blocking it keeps you in Google Search and its AI Overviews but out of Gemini app answers that ground in Search.

• Nothing

  Google Search and AI Overviews

  yes

  Gemini app grounding

  yes

  Gemini training

  yes

• Google-Extended

  Google Search and AI Overviews

  yes

  Gemini app grounding

  no

  Gemini training

  no

• Googlebot

  Google Search and AI Overviews

  no

  Gemini app grounding

  no (nothing indexed)

  Gemini training

  —

How do you get your pages into Gemini's answers?

The same way you get into Google, plus one decision:

• Be crawlable and indexed. Gemini's grounding draws on Google's index, so Googlebot must be able to fetch the page and Google must keep it. Check the page indexing report in Google Search Console.
• Decide on Google-Extended deliberately. Blocking it keeps you in Search but out of Gemini app grounding and future Gemini training. Allowing it does neither harm to Search.
• Write passages Gemini can lift. Grounding works on passages. A section that answers one question in its first sentences, with your brand and the qualifier inside, is what gets quoted.
• Be on the sources Google ranks for the question. If Gemini grounds in a review site or a comparison article, being on that page matters as much as your own.

What are the objections about Gemini and Google-Extended?

"We'll block the Gemini bot." There is no Gemini bot to block. Google-Extended is a usage control, read by Google's existing crawlers.

"Blocking Google-Extended keeps us out of AI Overviews." Google says it does not affect Search. AI Overviews follow Search's rules: indexed and snippet-eligible. Snippet controls are the lever there.

"Gemini and AI Overviews are the same." Same model family, same index, different surfaces and different controls.

"Our logs show no Gemini interest." Logs cannot show it. Read Gemini's answers instead.

How does Attensira track Gemini?

On Growth plans and above, Gemini is one of the engines Attensira can read every day, once you switch it on in settings: it asks your buyers' questions, stores Gemini's answers verbatim and records whether you are named and linked, with the sample size. The crew then works on what Gemini grounds in, your indexed pages and the sources Google ranks, and ships each change after your approval. Models lists the engines each plan reads, and Gemini visibility covers how a change there is measured.