// Ghost Citation Check

Are you the source without the credit?

Kevin Indig calls it a ghost citation: an answer engine uses your page, links it as a source, and never says your name in the text a reader sees. You supplied the answer. Somebody else got remembered. Enter your domain, your brand name, and up to three prompts. We put each one to ChatGPT, Perplexity and Gemini with web search live, then count, for each engine separately, how many of the answers that cited you failed to name you.

Write the brand name the way a journalist would write it in a sentence. We match it literally, so the trading name matters more than the legal one.

Prompts your customers ask (up to 3, checked on 3 engines each)

Use questions that do not contain your brand name. An engine asked about you by name will say it back, which tells you nothing about whether you earn the credit on a neutral question.

What a ghost citation is, and where the term comes from

Two things happen independently every time an engine answers. Your domain either appears in the source list or it does not, which is a citation. Your brand name either appears in the visible answer text or it does not, which is a mention. A ghost citation is the first without the second: the engine leans on your page and the reader walks away with the fact and no idea who supplied it.

The term and the measurement are Kevin Indig's, published in Growth Memo with Semrush across 3,981 domain appearances over 115 prompts, 14 countries and 4 engines. In that dataset 61.7% of citations were ghost citations, and only 13.2% of appearances earned both the link and the name. Those are his figures on his sample, not ours, and your own numbers will not look like them. Growth Memo.

Why this is counted per engine and never averaged

The same study found the engines behave differently enough to be treated as separate systems. ChatGPT cited in 87.0% of appearances and named the brand in 20.7%. Gemini ran the other way: named in 83.7%, linked in 21.4%. A single blended ghost figure across three engines describes none of them, so this tool never produces one. You get three sets of counts and the prompts behind each.

What this cannot tell you

Three prompts on three engines is nine runs, and some of those will not produce usable data. Engines regularly answer without searching the web at all. Those runs are thrown out of every count here rather than scored as a miss, because a run with no data is not a run where you were absent. In our own testing ChatGPT declined to search on roughly a third of runs, so expect to see the usable sample shrink.

What is left is a handful of observations. That is why every figure on this page is a raw count and not a percentage: a ghost rate computed from two citations would look like a measurement and behave like a coin toss. Read the result as a pattern worth looking at, not as a rate.

We also make no claim that closing a ghost citation gets you cited more often, or earns you revenue. Nobody has published data showing that, ours included. What the count tells you is where you are already doing the work of being the source without being known as it, which is a fact about your position, not a forecast.

Matching the brand name in answer text is literal. Case, spacing, hyphenation, possessives and a trailing suffix such as Inc or Labs are all handled. Nothing fuzzier than that, because a loose match invents mentions that were never there. If your brand name is also an ordinary English word, we say so and treat the ghost count as a floor, since an over firing name match can only push it down.

Different question from Citation Gap, which asks whether the engines cite you at all and shows which domains are cited instead. Start there if you are not sure you are in the answer yet. This is the free tier of Answer Monitor, which runs the same two states continuously.