Ghost Citation

A ghost citation is when an AI engine links to your domain as a source in its answer but never actually names your brand anywhere in the visible answer text — the citation exists, but the reader never sees who it came from. The term was coined by SEO strategist Kevin Indig and quantified in a joint study with Semrush (April 2026), which found that 61.7% of all AI citations across ChatGPT, Google AI Overviews, Gemini, and Google AI Mode fit this pattern — cited, but not named.

Key Takeaways

• A ghost citation is a citation (a source link) with no accompanying brand mention (your name in the answer text) — citation and mention are two separate, only weakly related events.
• The Semrush + Kevin Indig study (3,981 domain appearances, 115 prompts, 14 countries, 4 engines) found 61.7% of citations were ghost citations; only 13.2% of appearances got both a citation and a mention.
• ChatGPT and Gemini behave almost like opposites: ChatGPT cites in ~87% of appearances but names the brand in only ~20.7% of them; Gemini names brands in ~83.7% of appearances but cites (links out) in only ~21.4%.
• Comparative prompts ("X vs. Y") name brands roughly 2.4x more often than purely informational prompts.
• Standard AI-visibility tracking that only counts citations is missing more than half the real picture — citation rate and mention rate need to be tracked as separate metrics.

How Common Are Ghost Citations, Exactly?

Ghost citations account for 61.7% of all citations in the Semrush + Kevin Indig dataset — meaning for every 10 times a domain gets linked as a source in an AI answer, roughly 6 of those times the brand's name never actually appears in the text a user reads. Only 13.2% of domain appearances included both a citation and a named mention; the remaining 25.1% were mentions with no citation at all.

Citation and mention outcomes:

• Citation only (ghost citation) — 61.7% of domain appearances — linked as a source, brand name never appears in the text
• Citation + mention — 13.2% of domain appearances — linked as a source AND named in the answer
• Mention only — 25.1% of domain appearances — named in the answer, but no source link provided

Multiple secondary sources cite different headline numbers for the same phenomenon — 62%, 73%, "about 28%" — and these are frequently presented as if they contradict each other. They don't; they're measuring different things with different methodologies. The 61.7% figure from the original Semrush + Kevin Indig study is the one to anchor on here, since it's the source that publishes its full methodology (domain appearances as the denominator, not answers or clicks).

Why Do Different AI Engines Behave So Differently on Citations vs. Mentions?

Different AI engines cite and name brands at very different rates because of how each one is architected to generate answers, not because of anything a specific brand did right or wrong. ChatGPT tends to weave source links into flowing prose without narrating who wrote the underlying content, while Gemini more often names a brand conversationally without providing a clickable source link for it.

• ChatGPT: cites in roughly 87% of appearances, but names the brand in only about 20.7% — a citation-heavy, mention-light pattern.
• Gemini: names brands in about 83.7% of appearances, but cites (provides a source link) in only about 21.4% — the roughly opposite pattern.
• Prompt type matters too: comparative prompts ("Brand A vs. Brand B") produce brand mentions at roughly 2.4x the rate of purely informational prompts, since a direct comparison inherently requires naming the things being compared.

Why Does the Ghost Citation Gap Actually Matter for a Brand?

The gap matters because a citation without a mention provides almost none of the brand-recognition value marketers assume citations deliver — a user reading an AI answer that links to your page as source #7 in a footnote never learns your brand exists unless your name is actually written into the answer they're reading. Being "cited" and being "seen" are not the same outcome, even though most AI-visibility tracking tools report them as if they were.

This connects directly to why raw citation-count tracking is an incomplete measurement strategy for GEO — see AI share of voice for the broader measurement problem this fits into. A domain with a high citation count and a low mention rate can look strong on a citation-only dashboard while actually building almost no real brand recognition through AI search.

How Do You Reduce Your Own Ghost Citation Rate?

There's no confirmed, single fix yet, since this is a newly quantified problem without a mature playbook — but the practical, low-cost first step is separating the two measurements so you actually know your own rate, then working from there:

• Tag every tracked AI appearance as citation-only, mention-only, or both — not just "appeared" or "didn't."

• Calculate your own ghost citation rate (citation-only ÷ total appearances) per engine, since the pattern varies significantly by platform.
• Track the rate across multiple prompt cycles rather than a single snapshot, since AI citation patterns can shift 40–60% month to month per Semrush's tracking.
• Compare your rate against direct competitors on the same prompt set, not in isolation, since a "high" ghost rate only means something relative to what's achievable in your specific category.
• Prioritize consistent, distinctive brand naming in your own published content — the study's authors note that content structured with clear, repeatable brand attribution appears to correlate with better mention rates, though this remains an early, developing area of practice rather than a proven formula.

PERSONAL INSIGHT — PENDING: real anecdote goes here once about-page/resume detail is provided (e.g., a citation-vs-mention audit run for a client at Pyng or HCL). Leave as-is until real detail is supplied.

Frequently Asked Questions

Is "ghost citation" the same thing as the academic citation-fabrication problem researchers have been studying?

No — a February 2026 arXiv paper ("GhostCite") uses similar terminology for a different problem: LLMs fabricating invalid or nonexistent references in academic writing, which that paper found happening at rates of 14–95% depending on the model tested. The GEO/marketing usage on this page — citation without brand mention — is a separate, unrelated concept that happens to share a name.

Who coined the term "ghost citation" in the GEO/marketing context?

SEO strategist Kevin Indig coined it, publishing the first large-scale quantification of the phenomenon in a joint study with Semrush in April 2026, using data from Semrush's AI Visibility Toolkit.

Does a ghost citation still provide any SEO or GEO value if the brand isn't named?

It may still contribute to a domain's general visibility signal within an AI system's retrieval process, but it delivers essentially none of the direct brand-recognition value a named mention provides, since a reader has no way to know which source the linked citation actually is without clicking through.

Can a brand have a high citation rate and a low mention rate at the same time, and what does that mean?

Yes — this is exactly the pattern the study describes as a ghost citation problem, and it means a brand's AI-visibility dashboard may look strong on raw citation count while actually building little real-world brand awareness through AI search.

Do informational and comparative prompts produce meaningfully different ghost citation rates?

Yes — comparative prompts name brands at roughly 2.4x the rate of purely informational prompts, since directly comparing named entities requires the AI to state those names as part of answering the question at all.

Is the 61.7% ghost citation figure specific to English-language prompts, or does it hold across languages?

The underlying study covered 14 countries, suggesting the pattern holds at least directionally across multiple markets, though exact rates likely vary by country and language — check the primary study directly for any country-level breakdown before citing a specific non-US figure.

Do image or visual brand cues help close the ghost citation gap?

Some industry commentary building on the original study suggests consistent branded visuals (logos, consistent color and chart style, descriptive alt text) may help an AI system associate content with a brand even when the name isn't in the text — this is an emerging, not yet rigorously proven, practical recommendation rather than a confirmed finding from the core study itself.

Does Google Search Console or any standard analytics tool report ghost citation rate directly?

No — this requires dedicated AI-visibility tracking that separately logs citation events and mention events per appearance, which is not something traditional web analytics or Search Console currently reports.

Are ghost citation rates stable over time, or do they fluctuate significantly?

They fluctuate significantly — Semrush's broader AI Visibility tracking has found AI citation patterns shifting 40–60% month to month, meaning a single measurement snapshot is not reliable for tracking real progress; repeated measurement over multiple cycles is necessary.

Does a higher domain authority or larger content library reduce a brand's ghost citation rate?

The original study suggests domain authority and content volume appear to drive citation inclusion more than they drive being named — meaning a larger, more authoritative site may get cited more often without necessarily being named more often, which is part of why the two need to be tracked as separate signals rather than assumed to move together.