Brand mentions vs citations in AI answers: what's the difference

A brand mention in an AI answer is your name appearing in the text the reader sees: "For small teams, Brand A is the easiest option." A citation is one of your pages linked as a source for the answer, such as a numbered footnote, an inline link or an entry in a sources panel. They sound like the same thing, but they often don't happen together. In one large analysis, only 13.2% of brand appearances were both named and cited. The rest were one or the other. Mentions decide what the buyer hears; citations decide where the click goes and show you the evidence the engine used. Below: how to tell them apart, how each engine shows sources, how to measure share of voice against citation share, and what moves each one.

What counts as a brand mention in an AI answer?

A mention is any time the answer text names your brand or product, with or without a link. Mentions vary a lot in quality. Being the first recommendation ("the best choice for agencies is Brand A") is worth far more than being fifth in a list that ends with "other options include". A mention can also hurt you if it comes with a caveat like "powerful but expensive". That's why mention tracking has to record position and framing, not just whether you appeared. We cover the framing part in AI brand sentiment.

Mentions come from both of an engine's layers. The model can name you from training memory with no web search at all, or because the pages it just retrieved named you. Either way, a mention needs no link. Many AI answers name brands without citing anything; in our own scan data, most answers carried no citation at all.

What counts as a citation?

A citation is a link to a source page that the engine shows with its answer. Whether it counts as yours depends on what you measure, and there are really two kinds:

• First-party citations: a link to your own domain, such as your pricing page, docs or a comparison page you wrote.
• Third-party citations that name you: a link to someone else's page (a review site, a roundup, a Reddit thread) that itself mentions your brand. You get no click, but the engine's evidence is about you.

Most reporting, including Microsoft's, only counts the first kind. The second kind often matters more for being recommended, because engines cite third parties far more than they cite the brands they recommend.

How often are brands named without being cited?

Most of the time, by the best public data we found. Kevin Indig analysed 3,981 domain appearances across 115 prompts, 14 countries and four Google and OpenAI AI surfaces (data from Semrush's AI toolkit; also published in his Growth Memo). 74.9% of appearances were citations and 38.3% were mentions. Only 13.2% were both. That leaves 61.7% cited without being named, which he calls "ghost citations": the engine used the page but never said whose it was. Taking the overlap out of the mention figure gives 25.1% named without a link.

The engines behave very differently. In the same data, ChatGPT cited domains in 87.0% of appearances but named the brand in only 20.7%. It reads like an academic paper, with footnotes everywhere and few names. Gemini did the opposite, naming the brand in 83.7% of appearances and linking in 21.4%. Separately, BrightEdge reported in August 2025 that ChatGPT mentioned brands 3.2 times as often as it cited them. That's a different measure (brand names across all prompts, not appearances of particular domains), but it points the same way. Mentions and citations are separate events, and one engine can be strong on one and weak on the other.

How does each AI engine show its sources?

What counts as "cited" depends on the interface, so it helps to know how each engine displays sources:

Why do both mentions and citations matter?

Mentions are the recommendation. A buyer who reads "Brand A is the best fit for you" has heard your name with the engine's backing, whether or not they click. That matters a lot in categories where AI chats now shape shortlists. In G2's 2026 buyer survey, 69% of software buyers said they chose a different vendor than planned because of a chatbot. A citation without a mention (a ghost citation) gives the buyer your evidence with someone else's name on it, or no name at all.

Citations are the path and the paper trail. A citation is the only part of the answer that can send a visit. It also shows you which page the engine trusted, which is the most useful diagnostic you have. If competitors own the citations for your category, the answer is built from their framing. Citations tend to move first: a new page gets cited weeks before it changes what the model says from memory.

Share of voice vs citation share: how to measure each

The two need separate metrics, built from the same tracked prompts:

• Share of voice (mentions) = your brand mentions ÷ all brand mentions across the answers to your tracked prompts. It shows how much of the recommendation conversation is about you. Track position alongside it.
• Citation share (first-party) = citations to your domain ÷ all citations in those answers. Microsoft's version in Bing Webmaster Tools is defined per grounding query as "the percentage of citations attributed to your site out of all citations shown across all sites for that same grounding query", and Microsoft calls it observational, not a ranking.
• Evidence share (third-party) = citations to other sites' pages that name you ÷ all citations. This one is harder to collect, because someone has to read the cited pages, but it shows whether the engine's sources support you.

Worked example with made-up counts, to show the arithmetic - not data from any brand.

Reading the example: Brand A is named fairly often but rarely linked, while third-party pages about it are cited four times as often as its own site. That pattern suggests the reputation work is paying off and the first-party pages aren't, so the next job is making the owned pages more citable. The reverse pattern (high citation share, low share of voice) is the ghost-citation problem. For rolling these into one headline number, see what an AI visibility score is, and for the wider set of metrics, AI visibility KPIs.

What moves brand mentions?

Mentions follow what the web says about you as a whole. In Ahrefs' study of 75,000 brands in AI Overviews, the factor most correlated with being mentioned was branded web mentions (0.664), ahead of branded anchors (0.527) and branded search volume (0.392). All three are off-site. These are correlations, not proof of cause, but they agree with how the models work: an engine names the brands its sources keep naming. The work that moves mentions:

• Getting into the third-party lists, reviews and threads where your category is discussed (reviews, Reddit, roundups).
• A clear category sentence repeated everywhere, so the model files you under the right question.
• Patience with model updates: mentions from training memory change when a new model ships, not when you publish.

What moves citations?

Citations depend on whether your pages can be retrieved and quoted. The engine has to be able to fetch your page, find it for the search it ran, and lift a clean passage from it. Google says a page must be indexed and eligible to show a snippet to be a supporting link in AI Overviews or AI Mode. Microsoft's advice for Copilot is clear headings, tables, FAQs, current content and evidence. In practice:

• Let the engines' search crawlers in, and serve key content in HTML rather than client-side JavaScript.
• Give each common buyer question its own page or section that answers it in the first sentence.
• Keep facts fresh and say when they were updated. Use IndexNow and honest sitemap dates where they're supported.
• Put comparisons and prices in real tables and plain sentences, not images or tooltips.

How to turn ghost citations into mentions

If your pages get cited but your name doesn't make it into the answer, the engine is taking your facts and dropping your brand. In our experience (opinion, not a measured result), the fix is to put your name inside the sentence most likely to be quoted. Write "Brand A's 2026 survey found…" instead of "Our survey found…", and "Brand A starts at $29 per month" instead of "Plans start at $29". Original data, named benchmarks and clearly attributed definitions give an engine a reason to say who the source is. Generic explainers rarely do.

The takeaway

Mentions and citations are two scoreboards. A mention means the buyer heard your name; a citation means the engine used your page. Engines lean towards one or the other, and most appearances are only one of them. Track share of voice and citation share separately, per engine. Build third-party presence to win mentions and retrievable pages to win citations, and write your name into the sentences engines quote. Run a free Zene audit to see both on your own buyer questions: where you're named, and which domains the engines cite instead.