A citation in an AI-generated answer is not the same as a visible brand recommendation. Research into AI search results suggests that pages can be cited without the associated brand being named, while brands can also be mentioned without a linked or displayed source. For businesses tracking visibility in ChatGPT, Gemini, and other AI answer engines, those distinctions make citation totals an incomplete measure of performance.
The strongest takeaway is not that one source is universally more trusted than another. It is that citation, brand mention, and visibility are separate signals whose relationship changes across engines, prompts, and markets. The specific 2.55x gap raised in the originating discussion is not substantiated by the supplied primary research, so it should not be used as a benchmark for AI search performance.
Semrush's Ghost Citations Study provides a useful view of the issue. Published on June 9, 2026, the study found that about 62% of observed AI citations were "ghost citations": an AI system cited a page but did not mention the brand in its answer. About 25% were brand mentions without a citation, while roughly 13% included both a citation and a brand mention.
That pattern matters because a team may see growing citations while potential customers never see its name in the answer. Equally, a brand may be named in responses without receiving a citation that makes the underlying source visible or attributable. Neither outcome alone establishes actual user trust. They are practical indicators of how an engine presents information.
Why AI engines produce different visibility signals
Semrush's data points to sharply different presentation patterns between Gemini and ChatGPT. Gemini named brands in about 83.7% of appearances but cited sources only about 21.4% of the time. ChatGPT showed the reverse pattern in the study: it cited brands about 87% of the time but mentioned brands in only about 20.7% of appearances.
| AI engine | Brand mentions in observed appearances | Brand citations in observed appearances | What the pattern suggests |
|---|---|---|---|
| Gemini | About 83.7% | About 21.4% | Brand naming was much more common than source citation. |
| ChatGPT | About 20.7% | About 87% | Source citation was much more common than brand naming. |
These figures are not a universal ranking of the engines or a guarantee that a particular company will see the same results. They demonstrate why a single cross-engine score can obscure what users actually encounter. A citation-heavy result may support source attribution, while a named-brand result may improve recognition. A business seeking visibility should understand which outcome it is measuring before deciding whether performance has improved.
Additional analysis referenced in the supplied research reaches a compatible conclusion: there is no universal top source for brands in AI outputs, and cited domains can have limited overlap across engines. Cross-language work also indicates that third-party sources account for most brand citations, with market and context influencing the result. In practice, a strong presence in one engine, language, or query category may not transfer cleanly to another.
A practical measurement approach for AI search
For businesses, the immediate task is to replace a single "AI trust" metric with a more useful reporting framework. Brand mentions and citations can be treated as observable proxies for different outcomes, but neither should be presented as a definitive measure of authority, quality, or customer confidence.
A practical review should separate at least these questions:
- Is the company named? Track whether the answer explicitly identifies the brand, product, or website.
- Is a page cited? Record whether an AI engine links to or cites a page associated with the business.
- What is the answer context? Note whether the brand appears as a recommendation, an example, a comparison point, or a background source.
- Which engine and market produced the result? Keep engine, language, prompt category, and target market separate in reporting.
- Which third-party sources appear? Since third-party websites can drive citations, identify the publications, directories, reviews, and other sources repeatedly surfaced for relevant queries.
This approach makes research and SEO decisions more concrete. If a company is frequently cited but rarely named, it may need to assess whether its pages and external coverage make its brand association clear in the contexts AI systems use. If it is named but not cited, it can examine whether it has sufficiently useful, attributable material for the kinds of questions customers ask. The research does not establish a universal content fix, but it does show why treating all AI visibility as one number is misleading.
Measurement also needs consistent prompts. Teams should define the questions that reflect real customer research, then run the same prompt sets across relevant engines and markets over time. Changing prompts, geographies, or intent categories without labeling them can create apparent shifts that are really changes in the test itself. The aim is not to chase every mention. It is to understand whether AI answers surface the company in the customer journeys that matter.
AI search visibility can affect whether prospective customers encounter your company during research, but citation counts alone do not reveal the full picture. Scalevise can help you build a practical, engine-by-engine view of mentions, citations, and the questions shaping discovery through its AI Visibility and GEO Checker. Use that evidence to focus content and external visibility work where it can reduce blind spots and improve decision-making. Start an AI Visibility scan.
Frequently Asked Questions
What is a ghost citation in AI search?
A ghost citation is an instance where an AI system cites a page but does not mention the associated brand in the answer. Semrush found that about 62% of observed AI citations in its study fit this pattern.
Does a high AI citation count mean a brand is highly visible?
No. A cited page may not result in a visible brand mention. Semrush's engine-level findings show that citation and brand-mention patterns can differ substantially between ChatGPT and Gemini.
Is the reported 2.55x gap across AI engines confirmed?
No. The supplied primary research does not support 2.55x as a standalone finding, so it should not be used as a verified benchmark.
What should businesses track in AI-generated search results?
Track brand mentions, citations to owned pages, the context of each appearance, the engine used, and the relevant market or language. Reviewing these signals separately gives a clearer picture than a single visibility score.
Conclusion
AI-generated search results do not offer one universal measure of trust or visibility. The available research indicates that citations and brand mentions can diverge substantially, including between major AI engines. Businesses that separate those signals, test relevant prompts consistently, and account for market context will have a more reliable basis for improving how they appear in AI-assisted research.
