Ghost Citations: 62% of the Time, AI Cites You and Never Names You

By Eric Richmond, CiteMetrix

AI can use your page to build an answer without ever naming your brand.

That is not a rare edge case. It is the default in a significant share of AI citations, according to research from Semrush and Kevin Indig. In its September 21, 2026 article, “AI citation outreach: a Semrush & Claude workflow,” Semrush reported that 62% of AI citations produce no brand mention.

They call these ghost citations.

The distinction matters because a citation is not credit. A source attribution and a brand mention are two different events. Conflating them is one reason AI visibility reports can look healthy while brand preference does not move.

AI visibility is four different states

Most AI visibility discussions compress several distinct outcomes into one metric. In practice, there are at least four states:

1. Retrieved: An AI platform fetched or accessed your page.
2. Cited: Your URL appears as a source in the answer.
3. Mentioned: Your brand name appears in the answer.
4. Recommended: Your brand is positioned as a good choice.

These states form a progression, but one does not guarantee the next.

Your page can be retrieved without being cited. It can be cited without your brand being mentioned. Your brand can be mentioned without being recommended. And a recommendation can be positive, neutral, qualified, or even based on inaccurate information.

The last state is the one most directly connected to buyer behavior.

A source link may help an AI system construct an answer. A brand mention may put your name in front of the user. But a recommendation is where the system positions your company as a viable answer to the buyer’s question.

That is why “we are cited” is not the same as “we are winning AI search.”

The ghost citation problem

A ghost citation occurs when an AI platform cites a page that contains information about your category, product, or company, but does not name your brand in the generated response.

For example, an AI platform may cite:

• A comparison article that includes your product among many others
• A review that discusses your category but barely describes your company
• A community post that references your product indirectly
• A directory or marketplace page with incomplete brand information
• A third-party article containing outdated product details

The AI uses the page as evidence. It may absorb the page’s category framing, feature descriptions, and comparisons. But when it writes the answer, another brand receives the visible recommendation.

This creates an attribution gap:

• Your content contributes to the answer.
• Your URL receives the citation.
• Your brand receives little or no recognition.
• Another company may receive the recommendation.

The Semrush and Kevin Indig study is important because it separates source attribution from brand visibility. The 62% figure is Semrush research, not CiteMetrix research. It should be treated as an external benchmark, not as a CiteMetrix dataset.

Third-party pages can shape the answer, and get you wrong

The second-order problem is more serious than an absent mention.

The third-party page an AI platform trusts may describe your product incorrectly. Updating your own website will not automatically correct every external page that AI systems use as a source.

Semrush’s worked example makes this concrete. When Claude was asked to identify the best AI meeting notetakers, it recommended four brands that appeared repeatedly across comparison lists: Otter, Fathom, Granola, and Fireflies.

Claude also cited Medium reviews that contained outdated product information. Semrush specifically noted an example where Claude claimed Granola did not separate speakers, while Granola’s product documentation stated that speaker attribution was available through speaker tags on Zoom and Google Meet.

The lesson is not that one platform is uniquely unreliable. The lesson is that AI answers are assembled from an ecosystem of sources.

Your website is only one part of that ecosystem.

This is why AI visibility requires more than technical access and on-page optimization. It requires brand knowledge governance: the cross-functional work of keeping your company’s facts, naming, positioning, and public representations consistent across marketing, legal, product, and content operations.

How to diagnose your actual visibility state

The first practical step is to separate the signals instead of reporting them as one number.

Citation without mention: an attribution or entity-clarity problem

If AI cites your page but does not name your brand, investigate whether the model can clearly connect the page’s facts to your company.

Common causes include:

• Inconsistent brand naming across your website
• Product names used without a clear parent-brand relationship
• Pages that describe capabilities without identifying the company
• Weak organization and product structured data
• Third-party pages that mention your category but omit your brand
• Content that is useful as a source but unclear about who owns the information

This is where Source Pages analysis is useful. CiteMetrix shows exactly which pages, yours and your competitors’, AI platforms cite. You can identify the pages carrying your category’s answers, then determine whether those pages name you, describe you accurately, and position you prominently.

Mention without recommendation: an authority or positioning problem

Being named is progress, but it is not the finish line.

If your brand appears in an answer but is consistently described as a minor alternative, an expensive option, or a specialist for a use case you do not own, the problem is likely authority or positioning.

Look at:

• Where your brand appears in the answer
• Which competitors appear before you
• Whether the language is positive, neutral, or negative
• Whether your strengths are connected to the user’s actual question
• Whether the model presents you as a choice or simply as background context

CiteMetrix’s ModelScore™ Mention Score is designed to separate cited from recommended. It carries 45% of ModelScore, evaluates nine AI engines, and weights both sentiment and prominence rather than simply counting mentions.

The difference is important. Ten buried mentions are not necessarily stronger than two prominent, positive recommendations.

Absence: a signal-strength problem

If you are neither cited nor mentioned, the issue may be more fundamental:

• AI crawlers cannot access important content
• Your site lacks clear category and product information
• Your entity is weakly connected to the topics you want to own
• Competitors and third-party sources have stronger public evidence
• You have a Citation Gap for important queries

CiteMetrix Citation Gaps identify queries where your brand should be cited but is not. Pair that view with Competitor Share of Voice to see which named competitors are being recommended on each platform while you remain absent.

Then use Brand Perception and Model Sentiment to understand the language AI does use about you. This qualitative layer often exposes inaccurate third-party framing that a simple mention count would miss.

What to do about ghost citations

1. Improve entity clarity

Use one consistent name for your company, products, and key offerings.

Make the relationship between your organization and products explicit. State what you do, who you serve, and which category you belong to. Avoid forcing AI systems to infer basic facts from scattered or inconsistent language.

This is a brand knowledge governance responsibility, not just an SEO task. Marketing may own positioning. Product owns feature accuracy. Legal may approve claims. Content operations may maintain the public pages. The work succeeds when those teams use the same verified facts.

2. Put verified facts where AI already looks

Your own website is necessary, but it may not be sufficient.

Use Source Pages analysis to find the third-party pages that AI platforms repeatedly cite for your target queries. Then prioritize pages where:

• Your competitors are present and you are absent
• Your brand is mentioned but underrepresented
• Your product information is outdated
• A comparison or review misstates an important capability
• The page directly influences recommendation-stage prompts

The goal is not to manipulate every source. It is to improve the accuracy and completeness of the sources that actually shape answers.

3. Create a reliable source of truth

CiteMetrix’s Source of Truth turns verified Brand Facts into two deterministic outputs:

• An llms.txt file
• schema.org markup

These outputs are generated from your verified data, not written by AI. The same facts become the baseline used to check AI answers for accuracy.

CiteMetrix also uses a verified-only publishing gate. Only an Owner or Admin can verify facts for publication, which prevents unconfirmed information from becoming an official crawler-facing signal. Internal and live drift detection helps identify when the facts, generated files, or live site fall out of sync.

Explore the CiteMetrix Source of Truth to see how this works.

4. Monitor accuracy, not just visibility

A brand can be highly visible and still be represented incorrectly.

CiteMetrix’s Accuracy & Hallucination Detection checks AI responses against your verified Brand Facts. It identifies five error types:

• Incorrect Fact
• Outdated Info
• Fabricated Claim
• Misleading Context
• Critical Omission

Each issue receives one of three severity levels: Minor, Moderate, or Critical. Remediation is routed according to the error type, because correcting an outdated price is different from correcting a fabricated product claim.

This is brand-specific AI accuracy monitoring, not a general enterprise hallucination solution.

You can learn more about CiteMetrix Hallucination Detection.

5. Verify that the answer changed

Do not assume that fixing your page fixed the AI answer.

After you update your site or correct a third-party source, rerun the relevant prompts. Check whether:

• The source is still cited
• Your brand is now mentioned
• The wording is accurate
• Your prominence improved
• The recommendation changed
• The competitor still dominates the answer

CiteMetrix Correction Proof reports whether the AI platform actually updated its answer after a fix. That closes the gap between “we made the change” and “the model now reflects the change.”

The measurement argument

Ghost citations change what an AI visibility report needs to answer.

It is not enough to ask:

“Did an AI platform cite our domain?”

You also need to ask:

• Did the answer name our brand?
• Was the mention prominent?
• Was the sentiment favorable?
• Were we recommended?
• Which competitors were recommended instead?
• Which third-party pages shaped the answer?
• Was the information accurate?
• Did our correction change the answer?

That is the difference between measuring activity and measuring impact.

The CiteMetrix workflow is designed as a closed loop:

Monitor → detect → diagnose → fix → verify

Every other platform stops at the diagnosis. We don’t.

CiteMetrix brings Source Pages, Citation Gaps, Mention Score, Competitor Share of Voice, Brand Perception, Model Sentiment, Accuracy & Hallucination Detection, Source of Truth, and Correction Proof into one platform. Nine features work together across nine AI platforms on every plan, with everything included, no extra apps or add-ons.

Plans start at $79 per month.

See what AI says about your brand → citemetrix.com

Knowing that AI cited you is useful.

Knowing whether AI named, trusted, recommended, and accurately represented you is the real visibility advantage.