# AI brand mentions

> For the complete documentation index, see [llms.txt](https://attensira.com/llms.txt).
> Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.

> An AI brand mention is your name in the text of an AI answer, link or no link. How mentions are counted, why they differ from citations, with our own data.

- Canonical page: https://attensira.com/glossary/ai-brand-mentions
- Family: glossary
- Last updated: 2026-10-04
- Author: Karl-Gustav Kallasmaa (https://attensira.com/about)
- Claims as data: https://attensira.com/claims.json?slug=ai-brand-mentions

**In short:** An AI brand mention is your brand's name appearing in the text an AI assistant writes in reply to a question, whether or not the answer links to you. It is counted once per answer, separately from citations, because an assistant can name a brand it never links and link a page it never names.

| Figure | What it counts | n |
| --- | --- | --- |
| 197/198 | branded answers naming Attensira | 3 questions with our name, n=198 answers |
| 0/4,326 | unbranded answers naming Attensira | 58 other questions, n=4,326 answers |
| 63/4,326 | unbranded answers linking attensira.com | n=4,326 answers |

Window 2026-09-05 to 2026-10-04. attensira.com workspace, read over the Attensira MCP on 2026-10-04.

_Source: attensira.com workspace, read over the Attensira MCP, as of 2026-10-04 (n=4,524 answers)._

## Key facts

- Also known as: AI mention, LLM mention
- Unit: One answer, counted once however often named
- Mention rate: Answers naming you ÷ all successful answers
- Not the same as: [A citation: a link, counted separately](/glossary/source-citation)
- Read it beside: [Share of voice for your rivals](/glossary/share-of-voice)
- In the docs: [Measure › Share of voice](https://docs.attensira.com/measure/share-of-voice)

## What does an AI brand mention look like in real answers?

The fastest way to understand a brand mention is to look at a brand that is mentioned almost nowhere. Ours.

Attensira's own workspace for attensira.com reads 61 buyer questions on ChatGPT and Perplexity every day. Between 5 September and 4 October 2026 it stored 4,524 answers, and the strip at the top of this page is what we read from them over the Attensira MCP on 4 October 2026.

Three of the 61 questions contain our name: "what is Attensira", "Attensira reviews" and "Attensira vs Profound". Their answers named us 197 times out of 198. The other 58 are questions a buyer asks without our name, from "best GEO tools" to "how to get cited by ChatGPT". Their 4,326 answers named us zero times.

So every mention we have comes from a question that already contains our name. That is the normal state for a young brand, and it is why a single blended mention rate (197 of 4,524) tells you almost nothing until you split it.

What an earned mention looks like is visible on the same questions. Asked "check if ChatGPT mentions my brand", the engines named Ahrefs in 21 of 66 answers over the window, and Attensira in none.

## How is a brand mention counted?

A mention is your brand name appearing in the text the model wrote. No link is required. The unit is one answer: a name that appears eleven times in one answer counts once, the same as a name that appears once. That definition is in Attensira's [share of voice documentation](https://docs.attensira.com/measure/share-of-voice).

Detection reads the stored answer for the brand name as a whole word. It handles possessives, hyphenation and your domain written inside a URL, and it does not count "architecture" as a mention of a brand called Arc. What it cannot yet do is tell your brand from the same word used in ordinary prose. If your company is called Linear, "scales in linear time" counts. The error runs one way: common-word brands read high, never low.

## Why is a mention not a citation?

Because the model makes two separate decisions. It decides whether your brand belongs in the answer, and separately whether a page on your domain is worth linking. A [citation](/glossary/source-citation) is the second decision.

Our own counts show the two coming apart in both directions. On the three branded questions, 197 answers named us but only 176 of the 198 linked attensira.com, so 22 answers talked about us without pointing to a page of ours. On the 58 unbranded questions it ran the other way: 63 answers linked one of our pages and none of them wrote our name.

A report that collapses those into one number hides the only thing worth fixing. The first gap is a sourcing problem (the engine knows the name but not a page to quote). The second is an attribution problem (the engine quotes the page but not the name).

## Where does a mention come from?

Two mechanisms produce one, and they fail differently.

- **From memory.** The brand was common enough in [training data](/glossary/ai-training-data) that the model associates it with the category and writes the name without fetching anything. Nothing you publish this week changes it.

- **From retrieval.** A search step fetches pages, some enter the model's [context window](/glossary/context-window), and the model writes from them. Here your current page can be read, and the mention can be traced to a document.

Only the second responds to work you do now, and it has a hard prerequisite: the assistant's search agent has to be allowed to fetch you. OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, and that a [robots.txt](/glossary/robots-txt) change takes about a day to apply.

## What does an answer about brand mentions say when it does not name you?

This is one of 12 answers ChatGPT and Perplexity gave in 30 days to a question about tracking brand mentions. None of the 12 named Attensira. The five trackers ChatGPT did name are the ones whose own pages it linked, which is how a retrieval mention is built.

**ChatGPT, asked 2026-10-03:** "track brand mentions in ChatGPT, Claude and Perplexity"

> If you want to track brand mentions in ChatGPT, Claude, and Perplexity, you’re looking for AI visibility monitoring (sometimes called LLM mention tracking, AI search monitoring, or GEO tracking). These tools repeatedly run the prompts your customers ask, capture the AI answers, and measure whether your brand appears, how it is described, and which competitors are mentioned.

Sources cited, in order: analytics-agent.app, searchprism.io, mentionhq.com, visibrand.app, usebear.ai, linksii.com.

Attensira: not named.

_Source: attensira.com workspace, read over the Attensira MCP, as of 2026-10-04 (n=12 answers to this question)._

## What are the objections to tracking AI brand mentions?

**"Branded questions prove we are visible."** They prove the engine will repeat your name when it is asked about you. Our 197 of 198 sits beside 0 of 4,326 on the questions where nobody typed the name first. The name was in those answers because the question put it there.

**"A mention rate is a market share."** It is not. Every brand's rate shares one denominator, and one answer can name four brands, so rates do not add up to a whole. See [share of voice](/glossary/share-of-voice).

**"More mentions means more customers."** Nothing joins the answer a person read to a visit they made later. A mention measures presence in an answer. Revenue attribution is an inference you draw with your own analytics.

**"One reading is enough."** Models are not deterministic. The same question can be answered differently minutes apart, so a rate needs its sample size beside it before it means anything.

## How do you earn mentions on questions without your name?

You give the engine a reason to say your name on the pages it already reads. Those 63 unbranded answers that linked us and named nobody from Attensira are the clearest lead we have: the pages were read, and the sentences taken from them described what the product does without saying whose product it was.

The fix is editorial: put the brand inside the sentence that answers the question, so the quote cannot travel without it, and get named on the third-party pages the engines link for your category. In a workspace, Attensira's crew finds those read-but-unnamed answers, drafts the passage change, ships it after your approval, and then counts mentions and citations separately, with the sample size, so you can see which one moved. The [mention rate](/features/mention-rate) page shows how that number is reported.

## FAQ

### What is an AI brand mention?

It is your brand's name appearing in the text of an answer an AI assistant wrote, such as a ChatGPT or Perplexity reply. No link is needed. It is counted once per answer, however many times the name appears, and it is a separate measurement from a citation, which is a link to your domain.

### Is a brand mention the same thing as a citation?

No. Attensira's measurement documentation defines a mention as your brand in the answer text and a citation as a URL the model attached to your domain, and counts them independently. In our own workspace, read over the Attensira MCP on 4 October 2026, ChatGPT and Perplexity linked attensira.com in 63 of 4,326 answers to questions that do not contain our name and named us in none of them.

### How are AI brand mentions detected?

By reading the stored answer text for the brand name as a whole word, which handles possessives, hyphenation and the domain inside a URL. It cannot yet tell a brand from the same word used in ordinary prose, so a brand named after a common word reads high, never low, and its answers are worth reading by hand.

### Why am I named in almost every answer to questions that contain my name?

Because the answer repeats the question. An answer that says it knows little about your brand still contains the name, and it is counted as a mention. Branded questions measure what the engine says about you; only unbranded questions measure whether it thinks of you.

### Can mention rates for several brands be added up like market share?

No. Every brand's rate is divided by the same set of answers, and one answer can name four brands at once, so the rates do not add up to a whole. Attensira reports share of voice this way on purpose, so the number does not move when someone edits the competitor list.