How to rank in Gemini: what gets a brand named and cited in Gemini answers

What ranking in Gemini means

Ranking in Gemini means one of two outcomes in the Gemini app: your brand is named in an answer, or one of your pages is listed as a source for it. The Gemini app is Google's assistant at gemini.google.com and in the Gemini mobile apps. The Gemini Apps help page on sources says sources and related links can appear within and below a response, and that not every response includes them.

A mention and a citation are worth counting apart. A brand can be named with no link to its site, and a page can be cited in an answer that recommends a rival. The two respond to different work, so a test that merges them hides which work is paying off.

The name Gemini also covers the models behind Google Search's own AI features. In May 2025 Google said it was bringing a custom version of Gemini 2.5 into Search for AI Mode and AI Overviews in the US. Those features live inside Google Search and follow Search's rules, which is why the table below keeps them apart from the Gemini app.

The Search controls for the second row are covered in the guide to opting out of AI Overviews. The rest of this guide is about the Gemini app.

How Gemini builds an answer

Gemini builds an answer from two inputs: what the model learned during training, and, when it decides a search would help, content from the Google Search index. Google's crawler documentation calls the second input grounding, and describes it as giving the model content from the Google Search index at the moment a prompt is answered.

Google documents the grounding steps for its developer API, in the Grounding with Google Search guide:

1. The model reads the prompt and decides whether a Google Search would improve the answer.
2. If it would, the model writes one or more search queries of its own and runs them.
3. The model reads the results, writes the answer, and attaches citations that tie parts of the text to source URLs.

Google does not say that the Gemini app runs exactly this pipeline. Google also does not publish how many results Gemini reads for each query, or how it chooses among them. Two points still follow from the documented design. The queries are the model's own, not your buyer's exact words. An answer with no search behind it draws only on what the model learned in training, which the same guide describes as knowledge with a cutoff date.

Grounding can also fail quietly. The Google Cloud grounding documentation says a response can come back without source metadata, for example when the sources found were not relevant enough, and that such a response is not grounded.

Does ranking on Google help you appear in Gemini?

Ranking on Google helps a page appear in Gemini's grounded answers, because grounding draws on the Google Search index. A page Google has not indexed cannot be supplied to Gemini that way. Whether an indexed page may be used for grounding in the Gemini app is decided by one robots.txt token, Google-Extended.

Google-Extended is a permission, not a crawler: Google crawls with its usual user agents and reads the token as an instruction. According to Google's description of the token, it covers two uses at once:

• training future generations of the Gemini models behind the Gemini app and Google Cloud's Gemini API;
• grounding in Gemini Apps and in Grounding with Google Search on Google Cloud.

Google says the token does not affect a site's inclusion in Google Search and is not a Search ranking signal. A site that disallowed Google-Extended to keep its pages out of model training has therefore also kept them out of grounding in the Gemini app, while its Google rankings stay where they were. The Google Cloud grounding page states the same rule for its own platform: pages that disallow Google-Extended are not used for grounding there.

Being indexed makes a page eligible. Ranking for the right queries is the competitive part, and the right queries are the ones Gemini writes, not your head keyword. Google's AI optimization guide, written for AI Overviews and AI Mode, says those features are rooted in Search's core ranking and quality systems. Google has published no matching statement for the Gemini app.

Which sources Gemini cites

Gemini cites whichever pages it drew on for a given answer, and Google does not publish which sites, or which kinds of site, it cites most. What Google does document is the range of sources. The Gemini Apps help center lists public websites, files the user uploaded, and Google Workspace documents or emails when Workspace is connected. It adds that a long direct quote from a webpage gets a link to that page, and that an image from the web shows its source.

For web sources, grounding means the pool is whatever Google Search returns for the queries Gemini wrote. That pool can hold review sites, comparison articles, directories and forums as well as your own pages. A brand can therefore be described in a Gemini answer through someone else's page, and that page is the one Gemini lists.

The Gemini Apps help page describes a Sources panel of links. It does not describe any view of the searches behind them. The developer API returns both, which the method two sections below uses. The same approach for other engines is in the guides to getting cited in ChatGPT and getting cited in Perplexity.

How to check what Gemini says about your brand by hand

Checking Gemini by hand works when every run uses the same questions under the same conditions, so that a change in the answers reflects a change in Gemini rather than in the test. Two Google help pages set the conditions. Gemini can personalize answers using memory of your past chats, and its sources can include your own uploaded files and connected Workspace content.

1. Write 10 to 20 questions a buyer would type, in their words. Make about half of them category questions that do not name you, such as "best invoicing app for freelancers", and half questions that do, such as "is [your brand] good for agencies".
2. Fix the account. Use one personal Google account, signed in, with Memory turned off under Settings & help, then Personal Intelligence. Connect no apps or Workspace to it, and keep the same country and language each time.
3. Pick one surface, the web app or the mobile app, and stay on it.
4. Ask each question in a new chat, three times. One run is an anecdote; three show whether a mention is stable.
5. Log every run in the table below, including the domains listed when you open Sources.
6. Repeat on the same day each month with the same questions, and compare counts rather than single answers.

From the log, two rates matter: the share of runs that name you, and the share that cite your own domain. Track them apart. Do not ask Gemini why it left you out: the Gemini Apps help page on responses warns that Gemini can misrepresent how it works, including how it cites sources.

How to see the queries and sources behind a Gemini answer

The Gemini API shows the search side of a grounded answer: the queries and the sources. With the Google Search tool switched on, the Gemini API returns the queries the model ran and citations that tie spans of the answer to source URLs. On Google Cloud, the same data arrives as grounding metadata: webSearchQueries, groundingChunks for the sources, and groundingSupports linking text to sources.

1. Send each of your fixed questions with the Google Search tool on.
2. Copy the queries the model ran. These are the searches to rank for.
3. Run those queries in Google Search and note where your pages appear.
4. Check the cited sources for your own domain, and list the other domains.

Three cautions apply before relying on it:

• The API is not the Gemini app. It has no memory of your chats, and Google does not say the app uses the same model or settings.
• Each search is billed. On Gemini 3 models Google bills every search query the model runs; on Gemini 2.5 or older it bills per prompt.
• Showing grounded answers to other people in an application brings display rules: Google Cloud requires the Search Suggestions that come with a grounded response to be shown with it.

The overads Signals code handles two traps in this API, both in its source since July 2026. First, in the generateContent responses Signals reads, each web source arrives as a redirect link on Google's vertexaisearch.cloud.google.com host rather than as the cited page. The link has to be followed one step to learn the real page, and the code records that Google expires those links, so a stored link can stop resolving. Google's Cloud grounding page mentions redirect URLs only for image sources. Second, a grounded answer has sources by definition, so "the answer cited a source" says nothing about you. Count a citation only when your own domain is among the sources.

What to fix first for Gemini

The first fixes for Gemini are the ones that decide whether your pages can be used at all. Content work comes after them.

1. Read your robots.txt for Google-Extended. A disallow keeps those paths out of grounding in the Gemini app and out of future Gemini training. overads.io's own robots.txt, read on October 9, 2026, allows Google-Extended on its public pages and disallows it only on signed-in app paths such as /dashboard and /settings.
2. Confirm that the pages you want cited are indexed, with the URL Inspection tool in Search Console. Google's page on how to ask Google to recrawl your URLs covers requesting a crawl for a changed page.
3. Collect the queries Gemini writes for your buyers' questions, as in the previous section, and find out where you rank for each.
4. Answer those queries with something other pages do not have. Google's AI optimization guide asks for non-commodity content, such as first-hand experience, over restating what is already published. It is written for AI Overviews and AI Mode, which draw on the same index.
5. Read what the third-party pages Gemini cites say about you. Where they rank for Gemini's queries and describe you wrongly, or not at all, a corrected listing, review or comparison there can count for as much as a change to your own page.

The wider split between classic SEO and work aimed at AI answers is set out in GEO vs AEO vs SEO.

How long changes take to show in Gemini

Changes reach Gemini's grounded answers only after Google has recrawled the changed page. Google says crawling can take a few days to a few weeks, and that requesting a crawl does not guarantee the page will be included.

Changes to what the model knows without searching take longer. Google-Extended's description refers to training future generations of Gemini models, and Google publishes no schedule for when new training data reaches the Gemini app. A monthly check on fixed questions matches that pace. A result read a few days after an edit may not reflect the edit at all.

Where overads fits

overads is a marketing workspace where AI agents draft posts, publish them once approved and report on ads, and its Signals product is built to run the citation side of the API check in this guide on a schedule. Its AI-answer scan is switched off on the live product. As built, Signals puts tracked questions to Gemini and Perplexity, asks Gemini with Google Search grounding on, follows up to 10 of an answer's source links to the real pages, and counts a citation only when the brand's own domain is among the sources. It reads the sources Gemini returns, not the queries Gemini wrote.

• Signals is built for Gemini and Perplexity only. It does not cover ChatGPT, Microsoft Copilot, AI Overviews or AI Mode.
• Signals has no API endpoint or MCP tool, as the overads analytics API docs state, so its results stay in the overads app.
• overads does not edit robots.txt, request crawls or change how Google ranks a page. Those steps happen on your own site and in Search Console.
• A team that needs AI answer tracking running now, or across more engines, is better served by a dedicated tool. The comparison of overads and Peec AI sets out what one such tool covers.

The manual method in this guide needs no tool at all. A spreadsheet, a clean Google account and the same questions each month are enough to see whether Gemini names and cites you.