Semrush Analyzed 126 Million AI Prompts — 62% of Citations Are Ghost Citations That Never Name the Brand

Semrush just published the largest study ever conducted on AI search behavior. The company analyzed 126 million U.S. AI search prompts across ChatGPT, Gemini, Google AI Mode, and Google AI Overviews — a 50,400x expansion from its initial 2,500-prompt study released last year. The headline finding should alarm every brand investing in AI visibility: 62 percent of all AI citations are ghost citations.

A ghost citation means the AI engine includes your domain as a source — it links to your page, it used your content to generate its answer — but it never names your brand. The user sees a synthesized response that draws on your expertise, your data, your content, and your brand is nowhere in it.

This is not a bug. It is a structural consequence of how AI engines process content that lacks machine-readable brand identity.

The Numbers Behind the Ghost Citation Problem

The Semrush AI Visibility Index breaks down how differently each platform handles brand attribution, and the divergence is striking.

ChatGPT cites sources 87 percent of the time — it almost always provides links. But it names brands in only 20.7 percent of responses. That means for roughly four out of five citations, ChatGPT sends a link to your domain without ever telling the user who you are. Your content does the work. Your brand gets nothing.

Gemini inverts the problem: it names brands 83.7 percent of the time but includes citations — actual links to the source — in only 21.4 percent of responses. Your brand gets mentioned but the user has no path back to your site.

Neither platform delivers both. ChatGPT gives traffic without recognition. Gemini gives recognition without traffic. And across both, 62 percent of citations are ghosted — your domain appears in footnotes that users rarely read, while your brand name is absent from the answer itself.

On Gemini specifically, the mention-to-citation overlap drops as low as 30 percent. A brand can be named but not linked, or linked but not named, and the two rarely align.

Why Ghost Citations Happen

AI engines generate answers by synthesizing information from multiple sources. When a user asks "what is the best project management tool for remote teams," the engine reads dozens of pages, extracts relevant facts, compares attributes, and produces a synthesized answer.

The decision to name a brand in that answer — rather than merely citing its domain as a source footnote — depends on whether the engine can confidently identify the brand behind the content. This is where most brands fail.

Consider what the engine encounters when it reads a typical product page. The page has a logo image with a filename like logo-v3-final.png. The company name appears in the page title, the footer, and maybe an About Us paragraph. The product description is wrapped in marketing prose. None of this is structured as machine-readable identity data.

Now consider what the engine encounters when it reads a page with proper Schema.org Organization markup: a canonical brand name, a description, a logo URL, social profile links, contact information, and sameAs references that let the engine verify this entity across Wikipedia, LinkedIn, Crunchbase, and other knowledge sources. The engine can confirm the brand is a distinct, verifiable entity — and it uses that confidence when deciding whether to name the brand in its synthesized answer.

Ghost citations are what happen when the engine uses your content but cannot confidently resolve your brand identity. It defaults to a link because it can verify a URL. It omits your name because it cannot verify your brand.

Only 36 Brands Are Visible Everywhere

The Semrush study identified what it calls the "Universal 36" — the only brands that maintained top-100 visibility across all four AI platforms every month during the January-April 2026 study period. The list includes YouTube, Google, Reddit, Amazon, Facebook, Apple, Walmart, Disney, and Nintendo.

These brands share a common trait that has nothing to do with content quality or SEO optimization: they are entities that AI models can verify across hundreds of independent sources. Their brand identity exists in structured form across Wikipedia, Wikidata, Google Knowledge Graph, financial databases, government registries, and countless third-party references. When an AI engine encounters content from Apple, it does not need to guess whether "Apple" refers to a brand — the entity resolution is unambiguous.

For every other brand, entity resolution is a probability calculation. And when the probability is not high enough, the engine hedges by citing without naming.

The concentration numbers tell the rest of the story. In News and Media, the top three brands hold 82.9 percent of all AI visibility. In Consumer Electronics, it is 76.9 percent. These are near-monopoly conditions, and they exist because AI engines preference brands they can verify — which means brands with the densest structured identity data across the most sources.

In Finance, the top three hold only 41.4 percent. In Industrial, 42.2 percent. These are categories where the entity resolution landscape is less settled, where AI engines have not yet locked in their preferred sources — and where brands that build strong structured identity data now can claim disproportionate visibility before the window closes.

The Measurement Gap Makes It Worse

Ghost citations would be less damaging if brands could detect and respond to them. They cannot. Semrush found that 45 percent of marketing leaders cannot accurately measure their brand's visibility in AI-generated answers. Only 9 percent have tools to track metrics across all AI platforms.

A separate Searchless.ai analysis found that 88 percent of brands tracked were never mentioned in category recommendation queries — the high-intent queries where AI engines recommend specific products and services to users who are actively deciding what to buy.

This means most brands are simultaneously invisible in AI answers and blind to the fact that they are invisible. They continue investing in traditional SEO — optimizing for link-based rankings — while the discovery channel growing at 1,324 percent year-over-year goes unmonitored.

The traffic numbers compound the urgency. AI-referred visitors convert at 4.4 times the rate of traditional organic visitors. AI traffic to U.S. retail sites surged 1,324 percent between October 2024 and May 2026. Travel sector AI traffic grew 2,215 percent in the same period. And 93 percent of AI search sessions end without a click to any website — meaning the AI answer itself is the entire brand impression for most users.

When the answer is the impression, a ghost citation is not just a missed attribution. It is a missed brand interaction with a high-converting user who will never visit your site.

Earned Media Drives 84 Percent of Citations

One of the study's most consequential findings concerns where AI citations come from. Across the engines studied, 84 percent of cited links come from sources brands neither own nor pay for — earned media, editorial coverage, user-generated content, and community platforms like Reddit and Wikipedia.

ChatGPT relies heavily on Reddit and Wikipedia as source material. Gemini draws from a narrower pool that also includes YouTube. The citation frequency gap between brands with strong third-party trust signals and those without is 75x.

Twenty-seven percent of all citations are journalism-sourced. This means editorial coverage is not just good for brand awareness in the traditional sense — it is a direct input into whether AI engines can verify and name your brand.

The implication for structured data strategy is significant. Having Schema.org markup on your own website is necessary but not sufficient. AI engines cross-reference your self-asserted identity data against third-party sources. A brand that claims to be an authority on project management tools through its own structured data must also be referenced as such in earned media, industry publications, and community platforms.

The brands that avoid ghost citations are the brands whose identity can be verified across the sources that AI engines actually cite: earned media, community content, and editorial coverage — all pointing back to a structured, machine-readable brand identity on the brand's own domain.

What Brands Should Do Now

The ghost citation problem has a structural cause and a structural fix. AI engines ghost your brand when they cannot confidently resolve your identity. They name your brand when they can verify it across sources.

Build a machine-readable brand identity layer. Implement Schema.org Organization markup with your canonical brand name, description, logo, social profiles, contact information, and sameAs links to authoritative references (Wikipedia, Wikidata, LinkedIn, Crunchbase). This is the foundation that lets AI engines resolve your brand as a distinct entity.

Audit your cross-source consistency. AI engines cross-reference your brand data across your website, Google Business Profile, social platforms, directory listings, and third-party mentions. Every inconsistency — a different company name on LinkedIn versus your website, an outdated address in a directory — reduces the engine's confidence in your identity and increases the probability of a ghost citation.

Invest in earned media for AI citation. With 84 percent of AI citations coming from sources you do not own, your earned media strategy is now an AI visibility strategy. Focus on getting referenced in the source types each platform preferences: Reddit and editorial coverage for ChatGPT, YouTube and editorial for Gemini.

Monitor AI visibility across platforms. Citation rates vary 615x between platforms. A brand visible in Gemini may be ghosted in ChatGPT and absent from AI Overviews entirely. Track your brand's appearance — named versus ghosted versus absent — across all four major AI search platforms.

Close the measurement gap. Join the 9 percent of brands that can actually track AI visibility. If you cannot measure ghost citations, you cannot fix them. And with 54 percent of marketers planning GEO implementation within three to six months, the brands that measure first will optimize first.

Ghost Citations Are the Brand Data Problem Made Visible

The ghost citation phenomenon is the clearest evidence yet that brand visibility in AI search is fundamentally a data problem, not a content problem. Brands are being cited — their content is useful enough to be included — but they are not being recognized because their brand identity is not structured in a way AI engines can process.

This is not about writing better content. It is about building the data layer that makes your brand identity machine-readable, verifiable, and consistent across the sources AI engines actually consult.

The GEO market is projected to grow from 848 million dollars in 2025 to 33.7 billion dollars by 2034 — a 50.5 percent compound annual growth rate. Organizations unifying their SEO and AI visibility strategies report an 81 percent success rate for traffic and lead generation, compared to 36 percent for those managing them separately.

The ghost citation gap is closing. But right now, 62 percent of AI citations are still anonymous — and the brands that build structured identity data today are the brands that will be named tomorrow.

Find out if your brand is being ghosted. Extract your brand fingerprint and see what AI engines actually know about your identity.