# Mention vs Citation in AI Answers

> A mention means an AI answer names your brand. A citation means it links or attributes a source. They have different causes, so track both.

Page: https://trylectern.com/blog/mention-vs-citation-in-ai-answers

By Edgar Li, October 30, 2026.

A mention is when an AI answer names your brand in its text. A citation is when the answer points to a source: a link, a footnote, or an entry in a references panel. The two overlap less than most people assume. An answer can name you without citing anything, cite your page without naming you, or do both. Treating them as one number hides what is actually happening.

## Key Takeaways

- Mention: your brand appears in the answer text. Citation: a source is attributed, whether or not your name is in the prose.
- In [one study of 115 prompts across four engines](https://www.semrush.com/blog/the-ghost-citations-study/), 61.7 percent of brand appearances were a citation without a name, 13.2 percent were both, and 25.1 percent were a name only (Semrush, June 2026).
- Assistants differ. ChatGPT mostly cited without naming, Gemini mostly the reverse, in that study.
- Mentions travel with reputation and repetition across the web. Citations follow the pages an assistant retrieves.
- Track both, plus a third state: the answer names only competitors.

## Defining the Terms

**Mention.** The brand name, or a clear reference to it, appears in the generated text. "For small teams, Acme and Beta are common choices" is two mentions.

**Citation.** The answer attributes information to a specific source, usually a URL. Assistants show these differently: numbered footnotes, inline links, a sidebar or a list of sources at the end. A citation can be to a page you own or to a page about you on someone else's site.

**Recommendation.** The answer says to choose you. A mention is not a recommendation. [Scrunch (June 2026)](https://scrunch.com/blog/prompt-to-purchase-pipeline-how-ai-influences-buyer-behavior/) found that a recommendation does about twice the work of a mention in the browsing behaviour that followed, so it is worth recording separately.

Google's surfaces split further. In one AI Overview, your name can appear in the prose, your domain in the references panel, and a URL linked inline, and those three signals can disagree entirely.

## How Often They Come Apart

The [Semrush "ghost citations" study](https://www.semrush.com/blog/the-ghost-citations-study/) (June 2026) looked at appearances of a brand across 115 prompts in 14 countries and four engines. Of those appearances:

- 61.7 percent were a citation without the brand named,
- 13.2 percent were both named and cited,
- 25.1 percent were the name only.

The study also found that ChatGPT mostly cites without naming and Gemini mostly names without citing. A [second Semrush analysis](https://www.semrush.com/blog/chatgpt-topical-focus-study/) (August 2026) found that citations to a brand travel to categories far from its core, while mentions stay closer to home.

Two consequences follow. A tool that counts only mentions misses the pages that feed answers. A tool that counts only citations misses the answers where a person reads your name and goes looking.

## Why They Have Different Causes

**Citations follow retrieval.** When an assistant searches, it pulls pages and then cites a minority of them. In [AirOps's March 2026 analysis](https://www.airops.com/report/influence-of-retrieval-fanout-and-google-serps-in-chatgpt), ChatGPT cited 15 percent of the retrieved pages. In [its April 2026 report](https://www.airops.com/report/the-fan-out-effect-what-happens-between-a-query-and-a-citation), rank in the assistant's own results mattered most: pages in first position were cited 58.4 percent of the time and tenth position 14.2 percent. Access matters first, since an assistant can only cite a page it can read. See [AI crawler blocking](https://trylectern.com/blog/ai-crawler-blocking).

**Mentions follow the model's picture of your category.** That picture is formed from training data and from what many pages say. In a [large 2025 vendor correlation study](https://ahrefs.com/blog/ai-brand-visibility-correlations/), branded web mentions correlated with AI visibility far more strongly than backlink-based authority scores did. A brand repeated across independent sources is more likely to be named even when no single page is cited.

**Most citations are not to you.** Across studies, the brand's own site is a small share of what assistants cite on category questions, in some datasets in single digits: [4.3 percent of citations on category prompts](https://www.tryprofound.com/blog/enhanced-citation-categories) in one, [5.95 percent](https://scrunch.com/labs/citation-dynamics/) in another. Citations to third-party pages that mention you are what often put your name into an answer.

## What to Track

For each question in a fixed panel, record one of four states per assistant:

1. Named and cited (your page or a page about you).
2. Named, no citation.
3. Cited, not named.
4. Not named, with competitors named.

State 3 is the easy one to miss. It means an assistant used your content and did not credit you in the text. State 4 is the gap to close. Then track the rate of each over time, with the count of answers behind it. [AI share of voice](https://trylectern.com/blog/ai-share-of-voice) covers how to compare yours with competitors.

## Why It Changes the Work

If you are named but not cited, your problem is probably that assistants lack a page worth retrieving. If you are cited but not named, check how your pages describe the brand, since an assistant may be extracting a fact without the name attached. If you are neither, start with access and with where your category is discussed, not with your own pages alone.

Whatever you track, measure it over enough answers. One run is an example, as [how often AI answers change](https://trylectern.com/blog/how-often-do-ai-answers-change) explains.

Lectern records whether each answer names and cites your brand across ChatGPT, Gemini, Perplexity and Google AI Overviews, and the paid Lead plan adds Claude. Both outcomes are kept in the history so you can see which one moved. The [pricing page](https://trylectern.com/pricing) has the plans.
