61.7% of AI Domain Appearances Are Ghost Citations — How to Tell If Your Brand Is Actually Visible in 2026
61.7% of AI domain appearances in Semrush's study were ghost citations: the engine linked the domain but never said the brand name. Here's how to measure real AI visibility and fix it.
A ghost citation is when an AI engine uses your content as a source but never mentions your brand in the answer. Semrush's study with Kevin Indig logged 3,981 domain appearances across 115 prompts, 14 countries and four major AI engines and found that 61.7% of them were ghost citations: the domain earned a source link, but the brand name was absent from the answer (Semrush, June 2026; Indig first published the analysis in Growth Memo and Search Engine Journal in April 2026). Only 13.2% of appearances produced both a citation and a brand mention, and because 74.9% of appearances carried a citation link, about four in five cited appearances were ghosts. If you are measuring citations alone, you are measuring the wrong thing.
Your Content Is Working. Your Brand Is Not.
Most marketing teams track whether their pages show up as AI sources. That metric is misleading.
Seer Interactive's leading hypothesis, backed by six behavioral tests across 362,188 LLM responses, is that citations are post-hoc. The model generates its answer first — pulling brand names from parametric memory — then retrieves sources to support those choices. Your content gets appended as a footnote. Your competitor gets named in the answer (Seer Interactive, 2026). Seer calls this strongly supported behavioral evidence, not proven architecture.
The data signature supports it. In Seer's data, when a brand is mentioned in a response, that brand's citation rate jumps to 53.1%. When it is not mentioned, the citation rate drops to 10.6%. That is a 5x gap — and it runs in the wrong direction for anyone claiming citations cause mentions.
You are funding your competitor's recommendations with your own content.
Every AI Engine Ghosts Brands Differently
There is no single "AI visibility metric" because each engine treats citations and mentions differently:
| Engine | Cited (share of appearances) | Named (share of appearances) | Behavior |
|---|---|---|---|
| ChatGPT | 87.0% | 20.7% | Academic — cites sources heavily, rarely names brands in text |
| Gemini | 21.4% | 83.7% | Conversational — names brands often, rarely links sources |
| AI Overviews | 84.9% | 61.0% | Balanced — closest to equal citation and mention rates |
| AI Mode | 76.3% | 37.6% | Closer to ChatGPT's footnote style |
Source: Semrush ghost-citations study with Kevin Indig, June 2026, from the Semrush AI Visibility Toolkit: 3,981 domain appearances, 115 prompts, 14 countries. Each rate is the share of that engine's domain appearances.
ChatGPT dominates AI referral traffic: in Conductor's 2026 AEO/GEO Benchmarks Report (US data, May to September 2025), 87.4% of AI referral traffic, on average across 10 industries, came from ChatGPT. And ChatGPT is the heaviest ghost-citer in the Semrush data: at least 66.3% of its appearances were cited without being named (87% cited minus 20.7% named). If your strategy is "get cited by ChatGPT," you are optimizing for the engine that is least likely to say your name.
The Fix Is Not More Content
The instinct is always to publish more. Write another blog post. Create another comparison page. Optimize another landing page.
That instinct is wrong here. Ghost citations exist because the model's parametric memory — the knowledge baked into its weights during training — does not associate your brand strongly enough with the category. More self-published content on your own domain does not change parametric memory.
What changes parametric memory is what the model learned from before you asked it anything.
The brands that get named by AI engines are the ones third-party publications already write about. AirOps found that 85% of brand mentions originate from third-party pages rather than owned domains, and in a sample of 21,311 brand mentions across ChatGPT, Claude and Perplexity, brands mentioned for top-of-funnel commercial queries were 6.5x more likely to come from third-party content than from the brand's own site (one vendor's data).
Earned media placements in publications that AI engines already index and trust — Forbes, TechCrunch, Harvard Business Review, Wired — are the mechanism that writes your brand into the model's memory. Not more blog posts. Not more keywords. Third-party credibility in the places machines already read.
How to Measure Whether You Are Ghosted
Stop measuring citation count. Start measuring share of citation — your brand's share of all qualifying citations in the AI-generated answers a tracked query set produces. That is the metric that separates visibility from invisibility.
The measurement framework:
- Define 20-50 buyer-intent prompts across your category — branded, comparison, problem, and recommendation queries.
- Run each prompt across all five engines (ChatGPT, Perplexity, Gemini, Claude, AI Mode) monthly.
- Tag each appearance as citation-only, mention-only, or both. Citation-only is a ghost.
- Calculate share of citation per engine and aggregate. Compare to competitors across cycles, not single snapshots — in Profound's citation-drift study, 40.5% to 59.3% of the domains cited for identical prompts had changed a month later, depending on the platform.
- Track the ghost citation rate — the share of your appearances that carry a source link but no brand mention. Semrush's sample ran at 61.7%; the closer yours is to that, or above it, the more your content is working for everyone except you.
The cheapest version that works: 20 prompts, a spreadsheet, and 45 minutes a month. The output matters more than the tool.
The Architecture That Fixes This
Ghost citations are not a content problem. They are a citation architecture problem — the gap between being a source and being a brand in the model's memory.
This is what Machine Relations was built to address. The discipline starts with the same mechanism PR has always used — earned placements in trusted publications — but applies it to machine readers, not just human ones. When a prospect asks ChatGPT who leads your category, the answer is downstream of your editorial presence in publications that AI engines already trust. Not downstream of your blog output.
I coined Machine Relations because the data kept showing the same pattern: brands with strong earned media presence in high-authority publications appear in AI answers by name. Brands without it get ghost-cited — used as sources, never recommended.
AuthorityTech's visibility audit measures exactly this: where your brand shows up, where it is ghosted, and what the earned media gap looks like across all five engines. If your share of citation is not growing, your PR program is producing visibility that machines cannot use.
FAQ
What is a ghost citation? A ghost citation occurs when an AI engine uses your content as a source link but does not mention your brand name in the response text. Semrush's study with Kevin Indig found that 61.7% of 3,981 domain appearances across ChatGPT, Gemini, AI Overviews, and AI Mode were ghost citations (Semrush, June 2026). The measured unit is the share of all logged domain appearances that were cited as a source but never named; because 74.9% of appearances carried a citation link, that is about four in five of the cited appearances. This is attribution-behaviour evidence in an observed sample; it does not establish a universal ghost-citation rate, a provider's naming rule, citation causation, recommendation lift, pipeline, revenue, or any other business outcome.
What is share of citation? Share of citation is your brand's share of all qualifying citations in a tracked buyer-intent query set's AI-generated answers. Divide by the answers instead and you have citation rate. It is the Machine Relations equivalent of share of voice — the metric that tells you whether your investment in content and PR is producing machine-readable authority or human-facing noise.
Who coined Machine Relations? I coined Machine Relations in 2024 to describe the discipline of earning AI citations and brand recommendations through earned media in publications that AI engines trust. Machine Relations reframes traditional PR for machine readers — keeping the earned media mechanism and rebuilding everything around it that was broken.
How do you fix ghost citations? Ghost citations exist because the AI model's parametric memory does not strongly associate your brand with the category. The fix is earned media in trusted publications — not more content on your own domain. In AirOps' sample of 21,311 brand mentions, brands mentioned for top-of-funnel commercial queries were 6.5x more likely to come from third-party content than from the brand's own site.