If a potential customer asked ChatGPT to recommend a provider in your industry today, would your name come up? For 1,200 brands recently analyzed, the answer is no — not across all platforms, not consistently, often not at all. According to Semrush's AI Visibility Index 2026 study, which analyzed 126 million prompts in the United States between January and April 2026 across ChatGPT, Google Gemini, Google AI Mode and Google AI Overviews, only 36 brands maintain consistent visibility across all four platforms every month of the period. The other 1,200 disappear completely on at least one.
This isn't about ranking position. It's outright invisibility: your brand isn't named, isn't cited, doesn't exist in the answer. And most people managing marketing or content still don't know, because there are no official reports to track these citations. The only way to find out is to test manually, prompt by prompt, platform by platform.
01 What separates the 36 brands that dominate from the 1,200 that disappear
The 36 brands that maintain visibility everywhere — YouTube, Google, Reddit, Facebook, Amazon, Instagram, Apple, eBay, Walmart, Etsy, Yelp, TikTok, Microsoft, Netflix, Disney, Home Depot, LinkedIn, IMDb, Ford, Spotify, Indeed, Adobe, FDA.gov, Pinterest, Samsung, Toyota, NFL.com, MLB.com, Chevrolet, Marvel, Honda, Prime Video, NBA.com, Best Buy, USDA.gov and Nintendo — aren't just bigger or better known. They've built something structurally different: an identity uniquely recognizable by language models, a presence distributed across sources that AI considers authoritative, content that lends itself to being extracted and summarized without ambiguity.
The visibility gap within this group is enormous. YouTube tallies tens of millions of mentions across the entire dataset, Nintendo remains in the millions — a factor of thirty or more. But even Nintendo, last on the list, still appears on every platform every month. The other 1,200 don't.
The difference doesn't lie in product quality or traditional marketing excellence. It lies in how AI models manage to identify, distinguish and cite a brand when they need to construct an answer. If your name appears only on your site and a few other places, if your category isn't clear, if the content that mentions you is jargon-heavy or poor in comparable data, the model skips you because it doesn't have enough signals to trust.
02 Mention versus citation: two metrics that matter differently
The Semrush study distinguishes two ways a brand can appear in AI responses. A mention is when the model names the brand in the response text. A citation is when the model cites the brand's own site as a source. These are separate things, and their overlap varies enormously across platforms.
On Google AI Overviews, most brands mentioned are also cited as sources — mention and citation often coincide. On Gemini, the overlap is markedly lower: the brands named in the response and those used as sources are largely different groups. This difference radically changes strategy.
If you're mentioned but not cited, it means your brand authority holds — the model knows you exist and are relevant — but your site isn't considered a reliable source to construct the answer. If you're cited but not mentioned, as happens to Wikipedia, Medical News Today or Healthline, you're infrastructure: the model uses you to answer about others, but doesn't talk about you.
Medical News Today has a Source Surplus of 4.6x: it's cited as a source 4.6 times more often than it's mentioned as a brand. Wikipedia 4.3x, Healthline 4.2x, WebMD 4.0x. These are sites that AI models see as authoritative raw material, not as the subject of conversation. The opposite of what a brand that wants to be recommended needs.
03 How many sources each platform cites (and what it means for you)
The average number of sources a platform cites per response — the citation density — varies by a factor of five. In Semrush tests, ChatGPT cited an average of 15.4 sources per response, drawing heavily from Reddit and Wikipedia. Google AI Mode cited 11.4, with strong weight on Facebook, Instagram and Yelp. Google AI Overviews 9.2, dominated by YouTube as the single most important source. Gemini only 3.3, concentrating on YouTube, Wikipedia and e-commerce platforms.
If you're working on a platform that cites 15 sources, you have more slots available — and therefore more chance of getting in even if you're not among the very top in authority. If the platform only cites 3, you must be in the very high tier or you don't count. In many cases, the number of slots also varies between versions and updates of the same model. Less space, tougher competition.
This explains why the same brand can appear in most responses on one version of ChatGPT and few on another: the underlying model changes, citation density changes, and your brand slides in or out of the threshold. You don't control which model version responds, you don't control how many citation slots you have. You only control how authoritative you are relative to others competing for those slots.
04 The five diagnostic signals that predict invisibility
There are indicators you can measure today to understand whether you're on the trajectory of the 36 or the 1,200. These are things that already exist, only no one looks at them through this lens.
1. Mention-citation gap on your domain. If your brand is mentioned but your site is never cited as a source, you have a technical credibility problem: poorly extractable content, confusing structure, weak E-E-A-T signals. Test manually on ChatGPT and Perplexity with 10-15 queries in your sector. Count how many times you're named and how many times your domain appears among sources. If the ratio is below 1:3, you're in the red zone.
2. Mention-source overlap versus direct competitors. In manual tests, how many of the brands mentioned alongside you are also cited as sources? If overlap is high (above half) and you're rarely cited, it means the model considers your competitors more authoritative. If overlap is low (below a third), as on Gemini, the model is drawing sources from a different pool — probably general editorial sites or aggregators — and you're not there.
3. Source Surplus near zero or negative. If your domain is cited once for every ten times your brand is named, you have a Source Surplus below 0.1x. It means the model knows you exist but doesn't trust you enough to use you as a reference. Brands with surplus above 2x are those that AI uses as building blocks: your goal is at least to exceed 0.5x — cited at least half the times you're mentioned.
4. Extreme variability between platforms. If you appear in most responses on ChatGPT and few on Gemini, your authority is fragile. It holds when the model is generous with slots (15 sources), collapses when the model is strict (3 sources). The 36 universal brands appear on every platform every month, with contained differences.
5. Inconsistent brand entity across external sources. Search for your name on Google News, Reddit, industry forums. If you're called three different things, if your category changes from source to source, if half the mentions are wrong or ambiguous, the model can't construct a stable entity. Wikipedia uses "project management software," your site says "enterprise productivity platform," a review calls you "remote team tool." The model sees three different brands. You think you're just one.
05 How to test all this without dedicated tools
There are no official dashboards yet to track AI citations. Tools will come, but today the only reliable method is periodic manual testing. It works like this.
Write 15-20 realistic queries in your sector — ones a customer would actually make. Not "Who is [your brand]" (too obvious), but "What are the best [product category] for [customer type]" or "Compare [known competitor] with other options in the [your sector] market." Launch each query on ChatGPT, Perplexity, Google AI Overviews and Gemini. Note three things for each response: is your brand named? Is your domain cited as a source? Which other brands appear?
Repeat every month. If you notice mentions dropping, competitors gaining ground, your domain vanishing from citations while the brand remains named, you have a clear trend. Don't wait for total collapse. The slope corrects before it becomes free fall.
In our tests for clients operating in competitive verticals — B2B SaaS, strategic consulting, financial services — we regularly see brands with strong traditional SEO presence completely absent from AI responses. Not because Google penalizes them, but because the signals that matter to a language model are different from those that matter for ranking. A site can be first on Google for a transactional query and invisible to ChatGPT for the same question phrased in natural language.
06 Why AI invisibility doesn't fix itself with more content
The instinctive reaction when a brand discovers it's invisible is to produce more content. But volume without structure makes the problem worse. If you're already not being extracted correctly, publishing another ten articles in the same form changes nothing. The model continues not to trust, only now it has more material to ignore.
Brands with high AI visibility don't necessarily have more content. They have extractable content: paragraphs that answer a precise question, structured and comparable data (not "robust solution" but "supports up to 100 kg"), clear lists, sharp definitions, consistent author and date attribution. One article done this way is worth more than twenty written to "cover keywords."
And they have distributed presence on external sources that models consider authoritative. 85% of brand mentions in AI responses, according to recent research cited by industry analysts, comes from third-party pages, not owned domains. If you only talk about yourself on your site, the model has no confirmation. If you're discussed, compared, reviewed, cited elsewhere — on recognized publications, not anonymous forums — the model begins to trust.
This is the point where GEO (Generative Engine Optimization) separates from traditional SEO. SEO optimizes for ranking on a list. GEO optimizes for being cited inside an answer constructed by the model. The former requires relevance for a query. The latter requires authority perceived by the model through signals that are distributed, consistent and verifiable across multiple sources.
07 How to evaluate who can help you become visible
If you decide to work with someone on this problem, look for those who can show you concrete data. Real tests on prompts in your sector, monthly comparisons, brand-by-brand visibility. No one can guarantee a fixed position — models change without notice — but those working seriously know how to show you where you are today, where competitors are and which signals are missing.
Beware of those who promise "top 3 on ChatGPT in two weeks" or who sell content packages without first testing your current visibility. The trajectory builds over time, working on clear brand entity, presence on authoritative external sources, content structured for extraction. Done right, it shifts probability. But not in two weeks, and not with millimetric certainties.
Variability between models will remain. A brand that appears consistently on ChatGPT can disappear on Gemini because Gemini cites fewer sources and draws from a different pool. Aim to be strong enough to hold on the models that matter to your audience.
08 Frequently asked questions
How do I know if my brand is already invisible to AI without dedicated tools?
Launch 15-20 realistic queries in your sector on ChatGPT, Perplexity and Google AI Overviews. Note whether you're mentioned and whether your domain is cited as a source. Repeat every month. If you never appear or only appear on one platform, you're in the invisible tier. Manual testing is the only reliable method until official dashboards arrive.
Why does my brand appear on ChatGPT but not on Gemini?
Gemini cites an average of 3.3 sources per response, ChatGPT 15.4. Fewer slots available means tougher competition. If your authority is average, you hold on ChatGPT but vanish on Gemini. It's not that Gemini penalizes you: it simply has less space and chooses brands with stronger signals.
What's the difference between being mentioned and being cited as a source?
A mention is when the model names your brand in the response. A citation is when it cites your site among the sources used to construct it. You can be mentioned without being cited (the model knows you exist but doesn't trust your site) or cited without being mentioned (the model uses your content to talk about others). The second case applies to Wikipedia, the first to many B2B brands.
How long does it take to go from invisible to visible on AI?
It depends on where you start and how far you are from the 36 dominant brands. If you already have a strong SEO presence but are invisible to AI, it takes 3-6 months of work on brand entity, extractable content and presence on external sources. If you're starting from scratch, expect 9-12 months. Those who promise results in weeks are lying: models don't trust signals built quickly.
Can I use the same SEO content for GEO?
Only if that content is already structured for extraction: paragraphs that answer precise questions, clear lists, comparable data, without walls of dense text. If your SEO articles are optimized for keyword density and internal links but poor in structure, they won't work. Better to rewrite five well than to publish twenty in the same form.
Do the 36 dominant brands have anything in common beyond size?
Yes: unique brand entity, strong presence on external sources that models consider authoritative, content extractable without ambiguity, consistency between how they talk about themselves and how others talk about them. It's not just about budget: it's structure. An average brand can have these signals if they work on them, but it takes time and consistency.
How do you measure my brand's Source Surplus?
Divide the number of times your domain is cited as a source by the number of times your brand is mentioned in the text, in manual tests on 15-20 queries. If you're cited 3 times and mentioned 10, your Source Surplus is 0.3x. Below 0.5x means the model doesn't trust enough. Brands with surplus above 2x are those that AI uses as infrastructure, not as subject.
Why should I worry about AI Overviews if I already have traffic from organic Google?
Because AI Overviews are progressively covering organic results for informational and transactional queries. According to recent analyses, many traditional impressions lose value when the AI answer appears above: the user reads there and doesn't scroll. If you're not inside the AI answer, your organic ranking matters much less. It's like having a billboard behind a truck parked in front of it.
Overminds is an AI visibility and predictive growth agency, based in Milan. We continuously monitor ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews, Copilot, Meta AI, DeepSeek, Grok, Mistral and other generative engines to understand how they see and cite our clients' brands. If you want to know where you stand today and which signals are missing, request a free diagnostic analysis or download the self-assessment checklist to independently measure your domain's mention-citation gap and Source Surplus.