AI cites your page but never says your name. Which ghost citations actually cost you?

KEY TAKEAWAYS

1. Ghost citation rates are not comparable across studies. Check the denominator and the prompt mix before you compare.
2. Split your ghosts by what the answer does. Answers that name no brand are cheap. Answers that name rivals on your evidence are not.
3. Search Console and Bing Webmaster Tools count links, not names. You need your own prompt sampling to see ghosts.
4. Naming follows the model's memory and third-party coverage more than your own page copy.

A ghost citation is an AI answer that lists your page as a source but never says your brand's name. The term has spread fast this year, and so has a headline number: 62% of AI citations, or 40%, depending on which article you read.

Neither number describes your brand. The studies behind them count different things. And most ghost citations cost very little, while a small share cost a lot. This piece sorts out the numbers, separates the cheap ghosts from the expensive ones, and sets out what the evidence says gets your name into the answer.

Why do published ghost citation rates disagree so much?

Because each study divides by something different, asks different questions and samples different engines. Restated as a share of citations, the published figures look like this:

The Semrush figure is the one most often misquoted. Its 61.7% is a share of every appearance, including answers that named a brand without citing it. Of the appearances that were citations (74.9%), about 82% carried no name.

ChatGPT shows how wide the spread runs. BuzzStream's GPT-5.4-mini runs named 92.7% of the brands they cited. In the Semrush data, ChatGPT cited 87% of appearances and named 20.7%, so at least three in four of its citations were unnamed. Writesonic put ChatGPT at 37%.

The likeliest explanation is the prompt mix. BuzzStream asked about specific brands, and its results split sharply by prompt type. Single-brand prompts left 20.9% of citations unnamed. Head-to-head prompts left 34.7%. List and category prompts left 63.5%. Semrush's set mixed informational, how-to, comparative and commercial questions, and its domains included publishers such as medium.com, cited 16 times and never named. Informational prompts there produced a mention rate of 18%.

So the honest answer to "what is a normal rate" is that it depends on what you ask. Compute your own, on your own buyer questions.

Which ghost citations actually cost you something?

Start with what the link alone is worth. In Pew's March 2025 browsing panel of 900 US adults, people clicked a link inside a Google AI summary on 1% of visits. Readers mostly read the text. If your name is not in the text, the citation rarely reaches them. Seer Interactive did find that brands cited in AI Overviews had higher organic click rates on informational queries in 2025, 2.07% against 0.94%, but its authors say stronger brands are also more likely to be cited, so the link may not be the cause.

That does not make every ghost a loss. It depends on what the answer does instead.

• The answer names nobody. Problem questions often produce answers with no brands at all. Victorious found brands named in 0.10% of answers to problem-awareness prompts in its Q2 2026 study. Your citation here is the only credit anyone gets.
• The answer names your rivals. This is the expensive case. In Recrawled's 4 September 2026 run of 1,608 questions covering 40 UK software firms, eight firms had their pages cited in 19 answers that recommended other companies. Their content was used as evidence for a competitor's shortlist.
• The answer is a list that leaves you off. List prompts produce the most ghosts in BuzzStream's data. Being cited on a "best tools" question without being listed is close to the rival case.

There is also a reason to value the first kind. A Victorious report published on 26 October 2026, covering 161 brands and about 16,500 answers, found that 94% of brands cited on research questions were also named on buying questions, against 50% of brands that were not cited. Of 40 brands not named on any buying question in July, 81% of those cited on research questions had been named by October, against 21% of the rest. It is a correlation from a vendor, with small groups and no counts published, but it points one way. An unnamed citation on a research question may come before a named one on a buying question.

Should you care more about citations or mentions?

Credible people disagree. On 27 October 2026, Robert Rose at the Content Marketing Institute argued that citations matter more, because they show an engine relied on your expertise. Kevin Indig put the citation side plainly in April: "Being cited means an AI is drawing on your content."

The Recrawled and Victorious data argue the other way for vendors, because a buyer acts on the names in a shortlist.

Both are right for different jobs. Semrush's October 2026 visibility index shows how far apart the two signals sit. Fewer than half of the top 100 brands mentioned in a category were also among its top 100 cited domains, in every month from January to July. In July, ford.com ranked 18th for mentions in ChatGPT and 1,055th for citations. Investing.com ranked 50th for mentions and 7th for citations.

A research publisher or data source wants to be Investing.com: cited often, named sometimes. A product brand at the shortlist stage wants to be Ford. Decide which job each page does before you decide whether its ghosts are a problem.

What makes an engine say your name?

Mostly the model's memory and other people's pages, not your own copy.

Chen and Koudas tracked ChatGPT, Gemini and Perplexity daily for 90 days from 24 June 2026. When a cited third-party page named a vendor, the odds that the answer named that vendor rose 4.2 times for ChatGPT and 6.3 times for Gemini in a consumer market. In a niche B2B software market the figures were 39.7 and 17.7. A brand already in ChatGPT's memory for a question was named with 5.8 times the odds at equal citation support. And 46% of ChatGPT's mentions of brands the question did not name had no cited page naming them at all.

Other data agrees. AirOps found 85% of 21,311 brand mentions across ChatGPT, Claude and Perplexity came from third-party sources. Ahrefs' study of 75,000 brands found branded web mentions correlated with AI Overview mentions at 0.664, against 0.218 for backlinks.

The common advice is to put your brand name in the same sentence as your key claims. Writesonic's authors offer it as a hypothesis to test, not a finding. The mechanism is plausible. Claude's web search, for example, ties each citation to up to 150 characters of cited text, so the cited unit is a short passage, not a page. If that passage lacks your name, the answer built from it will too. It is cheap, so do it on your most-cited pages. Just do not expect it to beat what third-party coverage does.

In our terms, ghost citations touch three pillars of digital authority. AI visibility shows you are cited. Press coverage and brand entity decide whether you are named.

What does a ghost audit look like in numbers?

A hypothetical example, with made-up figures.

A software brand runs 40 buyer questions on ChatGPT, Gemini and AI Mode, for 120 answers. Its domain is cited in 50 of them and named in 20 of those. The raw ghost rate is 30 of 50, or 60%. That sounds bad.

Now sort the 30 ghosts by what the answer did:

• 18 were informational answers that named no brand.
• 9 named competitors and not this brand.
• 3 were category lists that left it off.

The costly ghosts are 12 of 50 citations, or 24%. The work list is those 12 answers. For each one, note which third-party pages were cited next to yours and which of them named the rivals. Those pages are your coverage targets. The 18 neutral ghosts go on a watch list, to see whether they turn into named answers on buying questions.

Which ghosts should you act on?

How do you measure your own ghost rate?

Platform reports will not show it. Bing's AI Performance report counts citations displayed as sources in Copilot and Bing AI summaries, and its Citation Share is your share of citations per grounding query. Google's generative AI report counts impressions as links to your site shown in AI Overviews and AI Mode. Neither counts names.

1. Write 30 to 50 buyer questions, split into informational, comparison and list prompts.
2. Run each on every engine you care about, several times. Answers drift, as covered in how to measure AI visibility when answers change.
3. Log citation and name separately for each answer. Record every brand named, not only yours. If you use APIs, read the inline citations, not the wider list of consulted URLs, which OpenAI says usually holds more links than were cited. Perplexity shows numbered citations, so check the text, not the source list.
4. Divide unnamed citations by citations, not by all appearances. State the denominator whenever you report it.
5. Classify each ghost using the table above.
6. List the third-party pages cited in the costly ghosts and which brands each names.
7. Rerun monthly on the same questions and track the costly ghost rate, not the raw one.

The short version

A ghost citation is proof an engine used your work. Whether it costs you depends on whose name the answer carries instead of yours. Count your own rate with a clear denominator, ignore the ghosts that name nobody, and work on the ones that name rivals. The lever with the best evidence is getting named on the pages engines already cite, which is what the placement marketplace is for.

Questions

What is a ghost citation in AI search?

It is an answer from ChatGPT, Gemini, Perplexity, Google AI Overviews or AI Mode that lists your page as a source but never names your brand in the answer text. The link proves the engine used your content. It does not put your name in front of the reader.

What is a normal ghost citation rate?

There is no single figure. BuzzStream found 30% of cited brand domains went unnamed, Writesonic about 40%, and Semrush data with Kevin Indig implies about 82%. The studies used different prompts, engines and denominators. Single-brand prompts produced few ghosts and list prompts produced many.

Are ghost citations bad for my brand?

Only some of them. An informational answer that names no brand at all costs you little, and may even lead to later recommendations. An answer that recommends competitors while citing your page as evidence is the expensive case, because a buyer reads the rivals' names and not yours.

How do I get AI engines to name my brand, not just cite it?

Get named on the third-party pages the engines already cite for your buying questions. In an October 2026 study, a cited third-party page naming a vendor raised the odds ChatGPT named that vendor 4.2 times in a consumer market and 39.7 times in a niche B2B market. Putting your name next to key claims on your own pages is cheap but untested.

Sources

1. Search Engine Land, Ghost citations. Why AI search cites your content, not your brand searchengineland.com
2. Semrush, Why 62% of AI citations don't lead to brand mentions semrush.com
3. Search Engine Journal, Kevin Indig, The ghost citation problem searchenginejournal.com
4. BuzzStream, How AI mentions and cites your brand (new study) buzzstream.com
5. PPC Land, ChatGPT, Gemini and Google never name 9 of 40 UK software firms in shortlists ppc.land
6. EIN Presswire, AI platforms nearly twice as likely to recommend brands they already cite, Victorious report finds einpresswire.com
7. Search Engine Journal, AI recognizes 96% of brands but mentions almost none, new study finds searchenginejournal.com
8. arXiv, Chen and Koudas, What drives AI citations? Content, retrieval and model knowledge in generative engine optimization arxiv.org
9. AirOps, Offsite brand mentions and compounding AI citations airops.com
10. Ahrefs, An analysis of AI Overview brand visibility factors (75K brands studied) ahrefs.com
11. PPC Land, Asda ranks 606th in ChatGPT citations as Tesco sits 25th, Semrush finds ppc.land
12. Content Marketing Institute, Your brand may be mentioned in AI answers, but is it cited? contentmarketinginstitute.com
13. Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results pewresearch.org
14. Seer Interactive, AIO impact on Google CTR. 2026 update seerinteractive.com
15. Bing Webmaster Blog, Introducing AI Performance in Bing Webmaster Tools public preview blogs.bing.com
16. Bing Search Blog, New AI visibility insights in Bing Webmaster Tools. Intents, topics, citation share, compare blogs.bing.com
17. Search Console Help, Generative AI performance report (Search) support.google.com
18. OpenAI, Web search tool guide developers.openai.com
19. Claude Platform Docs, Web search tool platform.claude.com
20. Perplexity Help Center, How does Perplexity work? perplexity.ai

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →