How to rank in Gemini: how Google's AI picks sources and brands
Gemini grounds its answers in Google's own index, so ranking in it depends on how well Google already understands your brand. Here is how it picks sources, what Google-Extended controls, and the steps that get you named and cited.
The seoscore team 10 min read
On this page
- What "ranking in Gemini" really means
- How Gemini decides which sources and brands to name
- Step 1: ask Gemini your customers' questions and note who it names
- Step 2: make sure Google can crawl and index the pages that matter
- Step 3: make Google understand your brand as an entity
- Step 4: write pages Gemini can lift an answer from
- Step 5: get named on the sources Gemini already cites
- Gemini, AI Overviews and AI Mode: what carries over
- Getting products into Gemini's shopping answers
- What doesn't help you rank in Gemini
- How to tell if it's working
- Track your Gemini visibility with seoscore
- Next steps
To rank in Gemini, Google has to be able to index your pages, understand who your brand is, and see you named on the sources it already trusts. The Gemini app grounds its answers in the Google Search index, so most of the work is SEO aimed at Google: crawlable pages, clear brand signals, pages that answer questions directly and genuine mentions on other sites. There is no submission form and no paid route in.
What "ranking in Gemini" really means
Ranking in Gemini is a form of GEO (generative engine optimization): getting your brand named and cited in the answers AI assistants write. Our GEO guide covers the idea in full. This article is about Google's assistant specifically.
"Gemini" covers more than one place. Most people mean the Gemini app and its web version at gemini.google.com, and Google also builds Gemini into other products such as Android and Chrome. Google Search has its own AI features too: AI Overviews (the summary above the results) and AI Mode (a conversational search mode). They draw on the same index, but they are part of Search and come with different controls for site owners. Keep the two apart when you measure results and when you change settings.
There are two outcomes to aim for:
- Being named. Gemini mentions your brand in the answer, for example in a list of recommended providers.
- Being linked. Gemini shows one of your pages as a source the reader can click.
You can get one without the other. A widely discussed brand can be named with no link to its site, and a useful guide can be linked while a competitor gets the recommendation.
GEO overlaps with two other disciplines. AEO (answer engine optimization) is whether AI engines can read, trust and quote a page. Classic SEO is whether Google crawls, indexes and ranks it. With Gemini the three are tightly linked, because they all run through Google.
How Gemini decides which sources and brands to name
Google hasn't published a ranking guide for the Gemini app. Its crawler documentation does describe grounding in Gemini Apps as "providing content from the Google Search index to the model at prompt time to improve factuality and relevancy" (Google's common crawlers page). That is the key point: when Gemini looks something up, it looks it up in Google.
Google's guide to generative AI features in Search describes query fan-out, where the model generates "a set of concurrent, related queries" to fetch more results (Optimizing your website for generative AI features). That guide covers AI Overviews and AI Mode, but the principle explains a lot. One question can become several searches, and a page that ranks for any of them can feed the answer.
Beyond ranking, a few Google signals are likely to shape who gets named:
- Entities. Google's Knowledge Graph holds facts about organisations, people and products (Google calls each of these an entity). If Google is confident what your brand is and what it does, it is easier to name you for the right question.
- Local data. For "near me" and town-specific questions, Google holds Business Profile and Maps data. Google hasn't documented how the Gemini app uses it, but a complete profile feeds Search and Maps either way.
- YouTube and forums. Google owns YouTube, and discussion threads rank widely in Google, so both can be among the pages Gemini draws on.
- What the model already knows. Brands that were widely written about before a model was trained can be named without any search at all.
Answers also vary. The model writes each answer afresh, fan-out searches can return different pages, and results differ by location and language. Ask the same question twice and you may get a different list, so measure over many runs, never one.
Step 1: ask Gemini your customers' questions and note who it names
Write 15 to 30 prompts in the words your customers use. Cover four types:
| Type | Example for a cycle shop |
|---|---|
| Category | What should I look for in a commuter e-bike? |
| Comparison | Is a hub motor or a mid-drive motor better for hills? |
| Best X for Y | Best e-bike for a 10-mile commute |
| Local | Where can I get an e-bike serviced in Leeds? |
Run each one in a fresh chat, ideally in a private window, so earlier conversations don't steer the answer. For each, record:
- whether you are named, and whether you are linked
- where you appear: first, mid-list or at the end
- which competitors are named
- which sources are linked
The sources list is the most useful part. It shows which sites Gemini trusts on your topic, and it becomes your target list for Step 5. Our guide to tracking brand mentions in AI assistants covers keeping these records. If you would rather not run them by hand, AI visibility asks Gemini the same kind of questions for you; more on that below.
Step 2: make sure Google can crawl and index the pages that matter
Gemini can't ground an answer in a page Google hasn't indexed. Google's own guide says its generative AI models "use publicly accessible, crawlable content".
Check indexing
- Open the Pages report in Google Search Console and read why important URLs are not indexed.
- Run URL Inspection on your key pages: the home page, your main services or products and your best guides.
- Check that robots.txt doesn't block Googlebot from sections you want cited.
- Check your main content is in the HTML rather than only appearing after JavaScript runs. See do AI crawlers render JavaScript? for why this matters.
Our technical SEO guide goes through these checks in more detail.
What Google-Extended does and doesn't control
Google-Extended is a product token you use in robots.txt. It is not a separate crawler and has no user agent string of its own. Google's documentation says it controls whether content Google crawls from your site may be used for training future Gemini models and for grounding in Gemini Apps. It does not affect your inclusion in Google Search and is not a ranking signal (Google's common crawlers page).
This is what a block looks like:
User-agent: Google-Extended
Disallow: /
One rule covers both uses. A site that refuses training also gives up its pages as grounding sources in the Gemini app. If you want Gemini to cite you, don't block it, and remove the rule if you added one earlier.
Google-Extended doesn't govern AI Overviews or AI Mode. For those, Google lists the usual snippet controls (nosnippet, data-nosnippet, max-snippet) and noindex (AI features and your website). Search Console also has a separate setting for Google's AI features in Search, which has no effect on training (Search Engine Journal). Google has also said it plans to support a shared AI-training token, __usage_ai-training, in robots.txt in 2027, and to treat an opt-out with it as a Google-Extended disallow (Search Central blog). The same trade-off will apply. For other AI crawlers, see how to allow AI crawlers in robots.txt.
Step 3: make Google understand your brand as an entity
Other assistants rely on their own search sources and training data. Gemini sits on top of Google's Knowledge Graph, so how clearly Google understands your brand matters more here.
- Organization markup. Add Organization structured data to your home page with your official name, URL, logo and
sameAslinks to profiles that are clearly yours. Google documents the supported properties on its Organization structured data page.
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Example Cycles",
"url": "https://www.example.com",
"logo": "https://www.example.com/logo.png",
"sameAs": [
"https://www.linkedin.com/company/example-cycles",
"https://www.youtube.com/@examplecycles"
]
}
- Google Business Profile. If you serve customers in person or in an area, complete it: categories, hours, services and photos, and reply to reviews.
- Consistency. Use the same name, address and one-line description on your site, social profiles, directories and press pages. Mixed versions make Google less sure they are the same business.
- Wikipedia and Wikidata. Only if you meet their notability rules. A deleted article helps no one.
- Knowledge Panel. If Google shows one for your brand, claim it through Google's verification process so you can suggest corrections.
Markup helps Google understand who you are, but it isn't a shortcut to citations. Read more in does schema markup help AI search?
Step 4: write pages Gemini can lift an answer from
- Put the question in a heading and answer it in the first two sentences beneath it.
- Use tables for comparisons such as features, prices and specifications.
- State facts with numbers, and say where each number comes from.
- Show a visible "last updated" date and keep it honest.
- Add an FAQ block for the follow-up questions customers really ask.
Google's guide says you don't need to rewrite pages specially for AI. These are habits that make a page clearer for every reader. The free AEO and GEO checker runs these checks on any page: AI crawler access, content without JavaScript, snippet rules, entity markup, freshness, citations, direct answers, FAQ markup and answer blocks. For the wider approach, see what is answer engine optimization?
Step 5: get named on the sources Gemini already cites
Go back to the sources you recorded in Step 1 and group them:
- Review and comparison sites. Make sure you are listed with accurate details, and ask happy customers to leave reviews.
- "Best of" articles. Contact the author with something they can check for themselves: a demo, a sample or data.
- Reddit and forums. Take part openly, as your brand, where you can help. No fake accounts.
- YouTube. Google owns it, so it is a natural source for Google's AI. Short videos that answer your customers' questions, with clear titles and descriptions, give Google more to go on. Independent reviews count for more than your own uploads.
- Industry publications. Offer expert comment, original data or a genuinely useful guest piece.
Google's guide warns that "seeking inauthentic 'mentions'" to influence what is said about your products "is not likely to be helpful", and that its generative AI features depend on both its ranking systems and its spam systems (Semrush's summary of the guide).
Gemini, AI Overviews and AI Mode: what carries over
| Gemini app | AI Overviews | AI Mode | |
|---|---|---|---|
| Where it appears | Gemini app and web, and inside Google products | Above Google's results | A conversational mode in Google Search |
| Where answers come from | Model knowledge plus grounding in the Google Search index | Google's index and ranking, with query fan-out | Google's index and ranking, with query fan-out |
| Site owner controls | Google-Extended in robots.txt | Snippet controls, noindex, Search Console AI setting | Snippet controls, noindex, Search Console AI setting |
| How to measure | Prompt tests, GA4 referrals | Search Console, rank tracking that checks AI Overviews | Search Console, prompt tests |
The work does triple duty. Indexing, entity signals, answer-ready pages and earned mentions help all three. Only the controls differ. For the Search side, see how to get cited in AI Overviews.
Getting products into Gemini's shopping answers
For product questions, Google relies on the product information it already holds about what you sell. You control two sources of it:
- Merchant Center. A product feed with accurate titles, prices, stock and identifiers such as GTINs.
- Product structured data. Markup on each product page with price, availability, ratings and reviews, as described in Google's product structured data documentation.
Keep the feed, the markup and the visible page in agreement. Google compares them, and mismatched prices or stock cause problems. We know of no Gemini-specific product guide from Google, so treat this as the same groundwork that powers shopping results in Search.
What doesn't help you rank in Gemini
- Stuffing prompts into pages. Lines like "best e-bike shop in Leeds" repeated down a page read as spam.
- Blocking Google-Extended to "force" citations. It does the opposite: it removes your pages from Gemini's grounding.
- Paying for placement. Anyone selling a guaranteed spot in Gemini's answers is selling mentions, which Google warns against.
- Relying on llms.txt. Google's guide lists it among things you don't need for its AI features. It is harmless, but it won't fix weak pages. See what is llms.txt?
- AI-generated filler pages. Thin pages written to catch questions rarely rank, so Gemini rarely finds them.
- Special AI versions of pages. Google's guide says content chunking, AI-specific rewrites and Markdown copies aren't needed.
How to tell if it's working
Re-run your prompt set once a month and work out two numbers:
- Mention rate: the share of prompts where you are named.
- Position: where you appear when you are named.
Because answers vary, run each prompt about three times and count how often you appear. One appearance is noise. Real change is a shift that holds for two or three months across several prompts, not one prompt flipping.
Also watch:
- GA4 referrals. Look for gemini.google.com as a referrer. Some visits arrive with no referrer and show as Direct, so treat the figure as a minimum.
- Search Console. Google counts traffic from AI features in your overall Search Console traffic.
- AI Overview citations. Rank tracking shows whether Google showed an AI Overview for your keywords and whether it cited your site.
The manual method gets slow fast. Twenty prompts, three runs each, across four assistants is 240 answers a month to read and log.
Track your Gemini visibility with seoscore
- Measure GEO. AI visibility asks Gemini, ChatGPT, Claude and Perplexity the questions your customers ask and scores how often, and how early, they name your site. You see every answer and the sources each one cites, which replaces the spreadsheet from Step 1. It is on every plan, Free included.
- Check your pages. Run the AEO and GEO checker on the pages you want cited, and fix what it flags.
- Watch Google's AI answers. On Starter and up, with a free trial, rank tracking checks AI Overview citations for your keywords, and AI mentions reports where Google AI Overviews and ChatGPT cite your site, which pages, the questions asked and the other sites cited beside you.
Next steps
- Write 15 to 30 customer prompts and run them in Gemini.
- Fix indexing gaps in Search Console and check robots.txt, including Google-Extended.
- Add Organization markup and complete your Google Business Profile.
- Rewrite key pages so each question gets a direct answer.
- Work through the sources Gemini cites and earn real mentions there.
- Re-measure monthly and look for changes that hold.
For the other assistants, read how to rank in ChatGPT and how to rank in Perplexity. For the bigger picture, start with what is generative engine optimization?
Questions people ask
How long does it take to start appearing in Gemini?
It depends on what is holding you back. Fixing an indexing problem or a Google-Extended block can matter as soon as Google recrawls the pages. Building entity signals and earning mentions on other sites usually takes months, so judge progress over two or three monthly checks.
Do I need separate content for Gemini, ChatGPT and Perplexity?
No. The same clear, well-sourced pages serve all of them. What differs is where each assistant looks things up, so the off-site sources worth targeting and the crawler rules you check vary between them.
Does Gemini always link to its sources?
Not always. Some answers come from what the model already knows and carry few or no links, while answers grounded in a search usually show sources. That is why being named and being linked are worth tracking separately.
Can a small or new brand get named in Gemini?
Yes, especially for specific or local questions where there are fewer strong answers. A new brand is unlikely to be named for broad category questions straight away. Start with the narrow questions your customers ask and the local signals Google already uses.
Sources
- Google's common crawlers developers.google.com
- Google's Guide to Optimizing for Generative AI Features on Google Search developers.google.com
- AI Features and Your Website developers.google.com
- Google Search Console AI Reports Rolled Out Worldwide www.searchenginejournal.com
- Supporting a shared AI training control for websites | Google Search Central Blog | Google for Developers developers.google.com
- developers.google.com/search/docs/appearance/structured-data/organization developers.google.com
- Google publishes guide to optimizing for generative AI search www.semrush.com
- developers.google.com/search/docs/appearance/structured-data/product developers.google.com
seoscore
Do AI assistants mention you?
Track how often ChatGPT, Claude and Gemini name your site, and what to change.
Get my free score