How AI decides which brands to recommend
When a buyer asks ChatGPT, Gemini or Perplexity what to use, the model hands back a short list of names. Here is the mechanism behind that list: how the question is expanded, which sources each engine reads, and the signals that decide who gets named.
By Raphaël Aubry, founder of Howseen AI · Updated October 6, 2026
3-day free trial, cancel in 1 click. Trial: 3 days, plans from €39 per month. Cancel anytime from your settings.
In short
The short answer
AI assistants do not rank pages the way Google does. They assemble an answer from the sources they already trust, then name the brands that appear consistently, are described clearly, and match what the buyer asked. To get recommended, you need to be present and unambiguous in those trusted sources, not just ranked on Google.
That is the whole game, and it is why a brand nobody has heard of can win a category answer while a bigger, higher-ranking competitor is left out. The rest of this page breaks down how the pick actually happens, step by step and engine by engine.
Under the hood
What happens when a buyer asks an AI
A recommendation is not a lookup in a ranked list. It is a synthesis, and it runs in four moves.
1. It expands the question. The model reads what you asked and quietly breaks it into sub-questions: who serves this audience, at this price, in this region, with these features.
2. It retrieves from what it trusts. It pulls passages from its training data and, for tools with live search like Perplexity, ChatGPT search and Google AI, from the current web.
3. It synthesizes and names. It writes a short answer and names the brands that show up most consistently across those sources and fit the intent best.
4. It attaches sources. Live-search answers cite the pages they leaned on, so you can see exactly which sources put a competitor in the answer instead of you.
The signals
What makes a model pick your brand
No single factor decides it. These are the signals that, stacked together, get a brand named.
Mentions in trusted third-party sources
Reddit threads, review sites, roundups and editorial coverage are cited heavily. Being named by others counts more than what you say about yourself.
A consistent, unambiguous description
When the same category, audience and value show up for your brand across the web, the model can quote you correctly. Mixed signals get you left out.
Answer-first, structured pages
Clear pages with one plain claim per section, an FAQ and schema give a model a clean sentence to lift. Walls of vague copy do not.
Fit with the exact question
You get named for the intents you actually cover. If a buyer question has no page and no mention behind it, a competitor fills the gap.
Freshness and specificity
Recent, specific and sourced content is quoted over stale or generic pages, especially by live-search tools.
Repetition across sources
One mention is noise. The same brand, described the same way, across several trusted places is the signal a model trusts.
The reframe
Google ranking and AI citation are decoupled
This is the part most teams get wrong. A page can sit on the second page of Google and still be the source ChatGPT cites, and ranking first guarantees nothing about being named. The two run on different signals, so being recommended by AI is measured and worked on separately from SEO.
The practical consequence: you cannot assume your SEO already covers this. You have to look at the AI answers directly, question by question, and work the sources behind them.
From mechanism to fix
What to do with the mechanism
This page explains how the pick happens. Acting on it is a separate job with its own guides: one diagnoses why a competitor is named instead of you and how to fix it, the other shows how to check what ChatGPT says about your brand today.
- →Diagnose and fix: why AI recommends your competitors instead of you, and what to change, on your site and off it.
- →Check today's answers: how to check if ChatGPT mentions your brand, question by question.
- →Measure it over time: what an AI visibility tool should track, engine by engine, and how to choose one.
Where Howseen fits. Howseen watches this mechanism for you: it tracks how ChatGPT, Gemini, Perplexity and Google AI talk about your brand, shows which sources get cited instead of you, and turns each gap into content you can publish. Start with what GEO is, or see GEO for SaaS, ecommerce and agencies.
The evidence
What the data shows about AI citations
Each AI engine picks its sources differently. Google's documentation and our dataset of 5,331 answers show why one brand gets cited and another does not.
- →Both AI Overviews and AI Mode may use a query fan-out technique, issuing multiple related searches for one question (Google Search Central). One buyer question becomes several searches, so you need answers on the related questions too.
- →In our study of 5,331 AI answers to 378 buyer questions (August 16 to October 4, 2026), YouTube is cited in 45.0% of Google AI Overviews answers and 27.4% of AI Mode answers, but in 0 of 1,164 ChatGPT answers. Reddit appears in only 0.9% of ChatGPT answers.
- →ChatGPT search finds pages through OAI-SearchBot, and OpenAI says sites that opt out of it will not be shown in ChatGPT search answers (OpenAI crawler documentation). More on this in our guide to AI crawlers and robots.txt.
- →The peer-reviewed GEO paper (Aggarwal et al., KDD 2024) found that optimizing content for generative engines can raise its visibility in AI answers by up to 40% (arXiv 2311.09735).
“Under the hood, AI Mode uses our query fan-out technique, breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf.”
Elizabeth Reid, VP, Head of Google Search, Google I/O 2025 announcement, May 20, 2025
The full dataset, engine by engine, is in who AI engines cite. This page is part of the GEO guide.
Questions, answered
How do large language models like ChatGPT decide which brands to recommend?+
It does not rank pages like Google. It assembles an answer from the sources it trusts, then names the brands that appear consistently across those sources, are described clearly, and fit the question. Being present and unambiguous in trusted third-party sources matters more than any single page you own.
Why does ChatGPT recommend my competitor and not me?+
Usually because your competitor is named in more of the sources the model trusts: review sites, community threads, roundups and well-structured pages. If those sources describe them for the exact question a buyer asks, and not you, the model repeats their name. For the step-by-step diagnosis and fix, read why AI recommends your competitors.
Does ranking first on Google get me recommended by AI?+
Not on its own. Search ranking and AI citation are decoupled. A page on the second page of Google can still be the source an AI cites, and ranking first does not guarantee being named. They are measured and worked on separately.
Which sources do AI engines cite most?+
It depends on the engine. In our study of 5,331 AI answers, YouTube was cited in 45.0% of Google AI Overviews answers and 27.4% of AI Mode answers, but in none of 1,164 ChatGPT answers, and Reddit appeared in only 0.9% of ChatGPT answers. Each engine builds its short list from its own sources.
How long does it take to get recommended by AI?+
It varies. Live-search tools like Perplexity and ChatGPT search can pick up a new source within days, while a model's baseline knowledge updates more slowly. Consistency across several trusted sources is what moves it, so it is a compounding effort, not a one-time fix.
Can I pay to appear in AI answers?+
Not directly today. The names in an answer are earned through the sources the model trusts, not bought as ad slots. The reliable path is visibility work: be present, clearly described and well sourced where the models look.
See how AI talks about your brand
Track your AI visibility across ChatGPT, Gemini, Perplexity and Google AI, and turn every gap into content that gets you cited.
3-day free trial, cancel in 1 click. Trial: 3 days, plans from €39 per month. Cancel anytime from your settings.
