How Gemini finds and cites brands
In short
Gemini answers from Google Search, so a site that Google Search can read is a site Gemini can find. Google also has a control called Google-Extended for its AI models, and Google says it does not affect Google Search. This page explains how Agentcites asks Gemini, how it shows its sources, and what a brand can do to be read by it.
How Agentcites asks Gemini
Agentcites sends each of your buyers’ questions exactly as written. Nothing of ours is added: no instruction to mention you, and no hint about what we hope to hear.
The question goes to a Gemini model through OpenRouter, a service that reaches many AI providers through one account. Web search is switched on, and it is Google Search, the same search Google’s own product uses. If that search is not available, the request is refused and recorded as a failure.
Every answer is stored word for word, with every link Gemini gave and the model that wrote it.
How Gemini uses the web
Gemini grounds its answers in Google Search results. Grounding means it looks up pages first and writes its answer from them, rather than from memory alone.
Its links do not point straight at the page. They pass through a Google redirect address, with the publisher’s site name given beside it. So Agentcites can say which site Gemini used, but often not which page on that site. Every screen that counts Gemini’s links says so.
Which robot reads your site
Google documents its robots on its common crawlers page. Two names matter here.
| Name | What Google says it controls | Its own visitor |
|---|---|---|
| Googlebot | Google Search, including Discover and all Google Search features | Yes |
| Google-Extended | Whether crawled content may be used to train future Gemini models and for grounding | No |
Google says Google-Extended has no visitor of its own. Google’s usual robots do the reading, and the Google-Extended name in robots.txt only acts as a control.
Google is clear about one thing: “Google-Extended does not impact a site’s inclusion in Google Search nor is it used as a ranking signal in Google Search.”
Google also says Google-Extended covers content used “for grounding” in Gemini Apps and the Vertex AI API for Gemini. How that applies to any one answer is Google’s to define. If being cited by Gemini matters to you, read Google’s page before you block Google-Extended.
What you can do to be cited by Gemini
- Make sure Googlebot can read the page. Gemini’s answers here are built on Google Search, and a page Google Search cannot see cannot be found that way.
- Read Google’s wording before blocking Google-Extended. It does not change Google Search, but Google says it covers grounding in Gemini.
- Answer the question in the page’s own words. The page Gemini grounds on has to say the thing it repeats.
- Keep your site name consistent. Gemini’s links often show only the site, so a recognisable name is what a reader sees.
- Watch which sites it uses, not only your own. If Gemini keeps grounding on a review site that does not list you, that site is the gap.
Questions
Does blocking Google-Extended hurt my Google ranking?
Google says no. Its documentation says Google-Extended does not affect inclusion in Google Search and is not a ranking signal. See Google's common crawlers page.
Why can I not see the exact page Gemini used?
Gemini’s links pass through a Google redirect address. Agentcites records the site named beside each link, but the page behind it is often hidden. We say so rather than guess.
Is this the same as Google AI Overviews?
No. AI Overviews appear inside Google Search and have no public way to ask them, so Agentcites does not measure them. This page is about Gemini, asked through its API with Google Search switched on.
Does Gemini give everyone the same answer?
No. The Gemini app can add a person’s history or location, and an answer can change between two identical runs. That is why Agentcites asks every week rather than once.