Gemini is the platform most businesses are ignoring and least able to afford to.
It went from under 9% of worldwide generative AI web traffic in June 2025 to around 27% to 28% by May 2026 (Similarweb, 2026 Generative AI Landscape Report), passed 750 million monthly active users (Google, Q4 2025 earnings), and shares a model family with AI Overviews, which reach 1.5 billion users monthly across 200+ countries (Google, Q1 2025 earnings).
So how does it actually choose which business to name? Unlike most AI platforms, Google documents this.
The documented mechanism behind Gemini's recommendations, why it rewards different things than Google ranking does, and the specific signals you can influence. Based on Google's own developer documentation rather than inference.
How does Gemini actually answer a question about businesses?
Gemini does not search the way you search. It writes its own queries first.
Google's documentation for Grounding with Google Search describes the sequence plainly: the model analyses the prompt, determines whether a search would improve the answer, and if so automatically generates one or multiple search queries and executes them. It then processes the results, synthesises them, and formulates a response with inline citations attached to specific text segments (Google).
Three consequences fall out of that, and they explain most of what looks strange about AI visibility.
Your customer's question is not the query that gets run. Someone asking "who should I use for bookkeeping in Brisbane" may trigger several generated queries about bookkeeping services, local providers, pricing and qualifications. You are competing for each fragment separately.
Ranking for the original phrase is not required. This is why Ahrefs found only 37.9% of AI Overview citations came from pages ranking in Google's top 10, down from roughly 76% nine months earlier, in an analysis of 4 million citations across 863,000 SERPs (Ahrefs). Our full breakdown is in the AI Overviews citation study.
Citations are attached to sentences, not to the answer. Google returns grounding chunks and grounding supports that map individual text segments to sources (Google). A page that cleanly answers one sub-question can be cited even if it is irrelevant to the rest of the answer.
Where does Gemini get local business information?
From a second grounding source that most AEO advice does not mention: Google Maps.
Google shipped Grounding with Google Maps as a distinct grounding tool alongside Search grounding. For a question with any local dimension, Gemini can ground its answer in Google's place data rather than in web pages.
That makes your Google Business Profile a direct input to Gemini's recommendations, not merely an SEO asset. Google's own documentation on how it sources information for Business Profiles confirms that profile data, user contributions and web information all feed local results.
The practical translation: for a local business, an incomplete Business Profile is not a missed SEO opportunity, it is a missing input to the answer engine. Yext's analysis of 6.8 million AI citations found 42% came from business listings and 44% from first-party websites, 86% combined from brand-managed sources (Yext).
Gemini can reach your business through Google Search grounding, which means your website, or through Google Maps grounding, which means your Business Profile. Most businesses work on the first and neglect the second. For local queries, the second often decides the answer.
What signals influence whether Gemini names you?
Four, in rough order of how much the available data supports them.
Branded mentions across the web. Ahrefs found branded web mentions carried the strongest correlation with AI visibility of any factor measured, at 0.664 (Ahrefs). Not backlinks. Mentions. How often credible sources talk about you by name.
Freshness. AI platforms cite content 25.7% fresher than what ranks organically (Ahrefs). Because grounding retrieves at query time, a recently updated page is more retrievable than an older one of similar quality.
Structured, self-contained answers. Since citations attach to individual segments, content that answers one question completely within a few sentences under a clear heading gives the model something it can lift and attribute. Sprawling prose that requires the whole page for context does not.
Consistent entity data. Name, address, phone, category and description need to agree across your site, your Business Profile and your major listings. Contradictory data makes an entity harder to resolve confidently, and confidence is what gets you named rather than described vaguely.
Is optimising for Gemini different from optimising for ChatGPT?
Less than you would expect. The retrieval mechanics differ, the underlying signals largely do not.
| Gemini | ChatGPT | |
|---|---|---|
| Retrieval | Google Search and Google Maps grounding | Web search plus training data |
| Local data source | Google Business Profile via Maps grounding | Web pages and third-party listings |
| Query handling | Generates its own search queries | Generates its own search queries |
| Citation rate | Varies by grounding mode | 6.8% of US prompts, up from 1.6% (Similarweb) |
| What earns citations | Branded mentions, freshness, clear answers | Branded mentions, freshness, clear answers |
The overlap is the useful part. Yext's finding that 86% of AI citations come from brand-managed sources holds across engines, and Ahrefs' branded-mentions correlation is not platform specific. Doing the work once lifts every platform, which is why we argue against running separate per-platform programmes in our AI search market share breakdown.
Where Gemini genuinely differs is the Maps grounding path. Nothing you do for ChatGPT improves your Business Profile, and for local queries that is the deciding input.
Why does Gemini sometimes recommend competitors who rank below you?
Because ranking and citation have separated, and Gemini is retrieving against generated queries rather than your target keyword.
The numbers make this concrete. 31.0% of AI Overview citations come from pages that do not rank in Google's top 100 at all, and 31.2% come from positions 11 to 100 (Ahrefs, March 2026). Across AI platforms generally, 80% of cited sources do not appear in Google at all (Ahrefs).
A competitor with a thinner site can be cited ahead of you if they answer the specific sub-question more directly, more recently, or with clearer structure. This is the same dynamic we cover in why your competitors appear in AI search and you do not.
There is a real upside to it. Concentration is high: the top 50 brands take 28.9% of all AI Overview mentions, and 26% of brands have zero mentions (Ahrefs). The field of businesses genuinely competing for citations is much smaller than the field competing for rankings.
What should you do about Gemini specifically?
Five things, in order.
Claim and complete your Google Business Profile. For local queries this is a direct grounding input via Maps. Categories, hours, services, description and photos all matter, and consistency with your website matters more than any single field.
Restructure key pages into question-shaped sections. Each heading a question, each opening paragraph a complete answer. This maps directly onto how grounding retrieves and attributes.
Update rather than accumulate. Given a 25.7% freshness gap, revising your best existing pages usually beats publishing new ones.
Earn mentions, not just links. Directories, industry publications, podcasts, partner sites, local media. The correlation says mentions move the needle harder than anything else measured.
Measure Gemini separately. If your reporting covers ChatGPT only, it misses roughly a quarter of the market and the fastest-growing platform. Our guide to checking your business in AI search covers doing this across engines.
Why Gemini and AI Overviews are not quite the same problem
They share a model family and a grounding system, but they meet users in different states of mind, and that changes what wins.
An AI Overview appears above search results for someone who typed a query into Google. They did not ask for an AI answer, they are often mid-task, and they may scroll past it. A Gemini conversation is deliberate. The user opened an assistant, and they frequently follow up, refine, and ask comparative questions across several turns.
That difference matters for the shape of your content. AI Overviews reward a clean, quotable answer to a single question. Multi-turn Gemini conversations reward content that survives interrogation: pricing that holds up when asked "is that good value", claims that stay true when asked "compared to what", and specifics that do not collapse when the user asks a second question.
| Google AI Overviews | Gemini conversation | |
|---|---|---|
| User intent | Typed a search, got an answer | Deliberately opened an assistant |
| Interaction | Usually one shot | Often multi-turn and comparative |
| Reach | 1.5B monthly users (Google) | 750M monthly users (Google) |
| Rewards | A clean quotable answer | Content that holds up to follow-ups |
| Local input | Search plus Business Profile | Search plus Maps grounding |
What Gemini cannot see about you
Worth stating plainly, because it sets the limits of the work.
Gemini has no access to your CRM, your reviews on platforms it does not crawl, your reputation in private communities, or anything a customer told a colleague. It grounds in what is publicly retrievable through Google Search and Google Maps at the moment of the query.
That has an uncomfortable implication. A genuinely excellent business with almost no public footprint is invisible to it, and a mediocre business with a well-maintained site, a complete Business Profile and frequent mentions is highly visible. AI visibility measures your public evidence, not your quality. Ahrefs' finding that 26% of brands have zero AI Overview mentions (Ahrefs) is largely a measure of how many good businesses have left no trace a retrieval system can find.
The fix is not persuasion. It is publication.
Does Gemini recommend you or your competitor?
Book a free AI Visibility Audit. We check Gemini, Google AI Overviews, ChatGPT, Perplexity and Claude, and show you exactly who gets named for the questions your buyers ask.
Book Your AI Visibility AuditFrequently Asked Questions
How does Gemini choose which businesses to recommend?
It generates its own search queries from the prompt, runs them against Google Search and, for local questions, Google Maps, then synthesises an answer with citations attached to individual text segments (Google's Grounding with Google Search documentation). Your business is selected per sub-question, not per query.
Does ranking in Google get me recommended by Gemini?
It helps, but far less than it used to. Only 37.9% of AI Overview citations come from pages ranking in Google's top 10, down from roughly 76% (Ahrefs, March 2026), and 31.0% come from pages outside the top 100 entirely.
Does my Google Business Profile affect Gemini?
Yes, for anything with a local dimension. Google offers Grounding with Google Maps as a distinct grounding source, so profile data can feed Gemini's answer directly. Yext found 42% of AI citations come from business listings.
Is Gemini worth optimising for separately from ChatGPT?
The content work is largely shared, since the signals that earn citations overlap heavily across engines. The one genuinely Gemini-specific investment is your Google Business Profile, which no amount of ChatGPT optimisation improves.
Why does Gemini recommend a competitor with a worse website?
Most likely they answer one of Gemini's generated sub-questions more directly, more recently, or with clearer structure. Citation is decided per fragment, so a narrower page that nails a specific question can beat a stronger page that buries the answer.
Ashur Homa
Built and scaled a digital brand to $100M+ in sales with zero ad spend. Has helped businesses generate millions through AI go-to-market strategy. Leads growth at Omni Eclipse.
Connect on LinkedIn
