How Gemini picks the brands it names, and how to track it.
Gemini names brands from what its model learned in training, and from Google Search when it decides a search will improve the answer. To be named you need to be described plainly on the public pages your category relies on, and readable when Gemini searches. Its answers change from run to run, and the sources it shows are not always the ones it used, so tracking Gemini means repeated questions, the competitors named beside you, and the pages behind each answer.
Does Proofsource track Gemini?
Yes, on Managed or a Custom plan: our team switches Gemini on after a call, and tracks it beside your competitors and its sources. Below is how Gemini chooses brands, from Google's own pages, and a method you can also run by hand.
- Gemini
- Tracked on Managed or a Custom plan. Our team switches Gemini on for your brands after a call, and it is asked the same buyer questions as every other engine you track. Book a demo.
- Self-serve plans
- ChatGPT, Perplexity, Claude and Google AI Overviews, on the free trial (25 questions, today and tomorrow) and on the price card. See how the four engines differ.
- More engines
- Google AI Mode, Gemini, Microsoft Copilot, Grok, DeepSeek and Meta AI are switched on by our team on Managed or a Custom plan, after a call.
How does Gemini decide which brands to name?
From what its model learned. Google says Gemini's models are pre-trained on a variety of data from publicly available sources. A brand that is described often and consistently on those pages is one the model can reach for without searching; a brand that is rarely written about, or easily confused with a bigger name, is not.
From Google Search, when it looks. Google says Gemini's responses may draw on external sources such as Google Search or one of its extensions, and that given a prompt it tries to retrieve the most pertinent information from them. Google's developer documentation for the Gemini API describes the grounded path step by step: the model decides whether a Google Search can improve the answer, writes one or more search queries, runs them, and writes a response grounded in the results, with inline citations. That documents the API, not the consumer app, but it is the clearest account Google publishes of how a Gemini model searches.
Differently every time. Google says a language model's response to the same prompt will not necessarily be the same every time. One check is one draw.
Google-Extended decides what it may use. Google documents Google-Extended as the robots.txt token that controls whether content it crawls may train future Gemini models and be used for grounding, and says it has no effect on inclusion or ranking in Google Search. Blocking it is a Gemini decision, not an SEO one; we cover the difference in blocking GPTBot does not remove you from ChatGPT.
Which sources does Gemini show, and can you trust them?
Google's help page says Gemini sometimes shows sources and related content within and below its response, that not every response includes them, and that the links may be content related to parts of the answer. A link under a Gemini answer is a lead, not a receipt.
Independent testing points the same way. When the Tow Center for Digital Journalism asked eight AI search tools to identify the source of exact article excerpts, more than half of Gemini's responses carried fabricated or broken URLs. We summarize the study, and why a citation is a claim the model writes, in zero citations does not mean unused.
How is Gemini different from AI Mode and AI Overviews?
All three run on Gemini models, and that is where the likeness ends for a tracker. AI Mode and AI Overviews are Search features built from Google's index; the Gemini app is a separate assistant with its own controls. A brand can be named in one and missing from the others for the same question.
| Dimension | Gemini app | AI Mode | AI Overviews |
|---|---|---|---|
| What it is | Google's assistant app, used on its own. | A conversational mode inside Google Search. | A summary above some Google results. |
| Where answers come from | What its model learned from publicly available data, plus Google Search or its extensions when it uses them. | Google Search, through query fan-out. | Google Search, which may use query fan-out. |
| Sources shown | Sometimes, within and below the answer; not every answer has them. | Links to supporting pages. | Links to supporting pages. |
| Control for your pages | Google-Extended, for training and grounding; it does not affect Search. | Googlebot and snippet controls (nosnippet, max-snippet, noindex). | Googlebot and snippet controls (nosnippet, max-snippet, noindex). |
| Tracked by Proofsource | On Managed or a Custom plan, switched on by our team after a call. | On Managed or a Custom plan, switched on by our team after a call. | On the free trial and every plan. |
For the Search side, see how Google AI Mode chooses brands and how AI Overviews choose brands.
How do you track whether Gemini names your brand?
- Use buyer questions that do not name you. "Best tools for" and "alternatives to" questions show whether Gemini reaches for you on its own; branded questions only test whether it describes you correctly.
- Ask from a clean context. Signed out where you can, with no saved preferences, the same wording each time, in the market you sell to.
- Record the competitors beside you. Whether you are named, in what position, and who is named instead. The competitors in the answer are the gap to close.
- Record the sources, then check them. Open each linked page and see whether it mentions you. A source that names your competitors and not you is the cheapest fix on the list.
- Repeat, and read a rate. 47% of day-to-day changes in our daily sample reversed the next day (half of daily changes reverse). Judge a change against several days, not one.
Proofsource runs exactly this on ChatGPT, Perplexity, Claude and Google AI Overviews, and on Gemini for Managed and Custom plans, with a 95% Wilson interval on each mention rate; see the methodology and how the four engines differ.
Three things that help in Gemini and in Search
One plain description
Say what you are in one sentence, the same on your site and your profiles, so a model can tell you apart.
Third-party coverage
Be on the public lists, reviews and comparisons your category relies on; they are what a model learns from.
A deliberate Google-Extended choice
Decide it on purpose: blocking it limits Gemini's use of your pages and does nothing for Search.
What people ask about Gemini and brands
Does Proofsource track Gemini?
Yes, on Managed or a Custom plan: our team switches Gemini on for your brands after a call, and records who it names beside you and which sources it shows. Self-serve plans track ChatGPT, Perplexity, Claude and Google AI Overviews.
Does blocking Google-Extended remove me from Gemini?
It limits Gemini, not Search. Google says Google-Extended controls whether content it crawls may be used to train future Gemini models that power Gemini Apps and the Vertex AI API for Gemini, and for grounding, and that it does not affect inclusion or ranking in Google Search. Blocking it can keep your pages out of Gemini's grounded answers.
Why does Gemini give different answers to the same question?
Google says a language model's response to the same prompt will not necessarily be the same every time, and Gemini may or may not search before answering. That is why one check is a sample, and why a mention rate needs repeated runs and an interval.
Are the sources Gemini shows the pages it used?
Not necessarily. Google's help page says Gemini sometimes shows sources and related content, that not every response includes them, and that the links may be content related to parts of the response. Read them as a lead to check, not as proof of where the answer came from.
Is Gemini the same as Google AI Mode?
No. AI Mode is part of Google Search and runs on a custom version of a Gemini model, but it answers from Search, using query fan-out. The Gemini app is a separate assistant that answers from what its model learned and sometimes from Google Search. Track them as separate engines.
- Google: Gemini Apps overview, how Gemini works gemini.google/us/overview/?hl=en, overview last updated Jul 25, 2024
- Google AI for Developers: Grounding with Google Search (Gemini API) ai.google.dev/gemini-api/docs/google-search, page last updated Oct 13, 2026
- Gemini Apps Help: view related sources and double-check responses support.google.com/gemini/answer/14143489?hl=en, undated help page
- Google Search Central: Google's common crawlers (Google-Extended) developers.google.com/search/docs/crawling-indexing/google-common-crawlers, page last updated Jul 14, 2026
- Google: AI in Search, going beyond information to intelligence (Elizabeth Reid, VP, Head of Search) blog.google/products/search/google-search-ai-mode-update/, May 20, 2025
Read on . Proofsource's own figures are linked to the post that reports each one, with its sample and its limits.
Related
See which AI engines name you today.
See which buyer questions name your brand, who gets named instead, and what to fix. 200 answers. 25 Prompts/Question. Top AI engines. 2 days. Free. No card.
Google AI OverviewsGoogle AI Mode
