AI Brand Mentions: What They Are and How They Differ From Citations

An AI brand mention happens when an AI-generated answer names your brand. If ChatGPT recommends Nike, Adidas and Brooks in response to a running-shoe prompt, each brand has received a mention.

It sounds straightforward. But counting mentions alone misses a lot.

Where your brand appears in the answer matters. So does what the engine says about you, which competitors appear next to you, and whether the sources behind the answer mention your brand at all.

And mentions should not be confused with citations. A mention is part of the answer the user reads. A citation is a source the AI engine surfaces or uses when constructing that answer.

Understanding that difference is the starting point for measuring brand presence in AI search properly.

What is an AI brand mention?

An AI brand mention is any instance where a brand is explicitly named in an AI-generated response.

For example, ask:

What are the best CRM platforms for a small sales team?

An answer might include:

HubSpot is a popular option for smaller teams, while Pipedrive offers a more sales-focused interface. Salesforce is usually better suited to companies with more complex requirements.

HubSpot, Pipedrive and Salesforce have all received brand mentions.

No link is required.

That makes AI mentions different from many traditional search metrics. A user can discover a brand directly inside ChatGPT, Gemini or another AI engine without visiting the brand's website.

The answer itself becomes part of the discovery process.

Not every mention has the same value

A simple mentioned/not-mentioned metric is useful, but it doesn't tell you how much exposure each brand received.

Consider:

Nike is a strong option for everyday running, with a broad range covering both training and racing. Adidas also has several popular performance models. You could also consider Brooks.

All three brands are mentioned.

But Nike appears first and receives the most attention. Adidas gets a shorter description. Brooks is briefly added at the end.

A useful analysis therefore needs to look beyond the number of mentions and inspect things such as:

• >whether the brand appears at all
• >where it appears in the answer
• >how often it appears
• >how it is described
• >whether it is actively recommended or simply listed
• >which competitors appear alongside it

This is one reason Vercite stores the full response alongside the underlying metrics. A number tells you that something happened. The answer shows you what actually happened.

How many brands does AI actually mention?

There is no standard number.

Vercite analysed more than 234,000 responses from ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode between March 1 and July 13, 2026.

During the final 14 days of that dataset, the average number of brand mentions per response was:

Brand mentions per responseDifferent brands per response

ChatGPT

4.5 3.3 · 1.4x repeats

Gemini

3.5 2.7 · 1.3x repeats

Google AI Overview

3.5 2.6 · 1.4x repeats

Google AI Mode

3.2 2.5 · 1.3x repeats

Perplexity

2.7 2.2 · 1.2x repeats

All figures are averages over the last 14 days (Jun 30 – Jul 13, 2026). “Repeats” = brand mentions divided by different brands.

The difference between the two numbers matters.

ChatGPT, for example, averaged 4.5 brand mentions but only 3.3 different brands per response. Some brands were mentioned more than once.

Across all five engines, the average answer contained only around two to three different brands during that period.

For commercial prompts, that can make inclusion meaningful. If someone asks for five project-management tools and the answer names three companies, being absent means the user may never encounter your brand during that stage of research.

AI brand mentions vs citations

Mentions and citations describe two different layers of an AI answer.

A mention is a brand name in the generated response.

A citation is a page or domain surfaced as a source for that response.

For example, an AI answer could say:

Nike Pegasus, Brooks Ghost and ASICS Gel-Nimbus are good options for everyday running.

The answer might then cite:

• >runnersworld.com
• >reddit.com
• >youtube.com

Nike, Brooks and ASICS received the mentions.

Runner's World, Reddit and YouTube received the citations.

The brands appearing in the answer and the domains appearing in the source list do not need to match.

This distinction is important enough that mentions and citations should be measured separately.

A mention answers:

Did the user see my brand?

A citation answers:

Which sources surfaced around this answer?

You usually want to understand both.

The four mention and citation states

Looking at mentions and citations together creates a useful diagnostic framework.

For a given prompt, your brand can broadly fall into four states:

These situations tell you different things.

1. Mentioned and cited

Your brand appears in the answer and content from your own domain also appears among the sources.

You have presence at both levels.

The next questions are about quality: where were you mentioned, how were you framed, and which of your pages surfaced?

2. Mentioned but not cited

Your brand appears in the answer, but your own domain does not appear among the citations.

Third-party pages may instead be contributing information about your category or brand.

That is common enough that it should not automatically be treated as a problem. A customer may learn about your company through a recommendation assembled from publishers, forums, review sites and other third-party sources.

The useful question is what those sources say.

3. Cited but not mentioned

Your website appears as a source, but the generated answer doesn't name your brand.

This is particularly interesting enough that we've given it a dedicated breakdown: why your domain gets cited but your brand doesn't.

It means your content surfaced at the source level without translating into brand exposure in the final answer.

You might see an even stronger version of this with third-party sources: an engine repeatedly cites a comparison article that discusses your competitors but leaves your brand out entirely.

That creates a clear place to investigate.

4. Neither mentioned nor cited

Your brand does not appear in the response and your domain does not appear among its sources.

One occurrence does not tell you much. AI answers vary.

But if the same pattern appears repeatedly across an important group of prompts, while competitors consistently appear, you have found a visibility gap worth investigating.

Citations can help explain brand mentions, but they are not attribution

This distinction deserves some care.

If an AI answer mentions your brand and cites five pages, you cannot automatically say that one of those five pages caused the mention.

AI engines can combine information from retrieved web pages, model knowledge, previous context and multiple searches. The exact process also differs between engines.

Citation analysis is therefore better treated as evidence about the source environment around an answer.

Vercite captures the sources alongside each tracked response and lets you inspect:

• >which pages were retrieved
• >which pages were ultimately cited
• >which brands those pages mention
• >which sources recur across prompts
• >where competitors appear but you do not
• >how citation patterns change between engines

This gives you somewhere to investigate rather than pretending AI attribution is more precise than it is.

AI engines mention brands differently

There is another problem with measuring AI brand mentions: there is no single version of an AI search result.

Ask the same commercial question in ChatGPT, Gemini and Perplexity and you may get different brands.

Vercite's 234,000-response study illustrates this.

During the final two weeks of the research period, ChatGPT named 3.3 different brands per response on average. Perplexity named 2.2.

The differences over time were even more pronounced.

Google AI Mode reached 8.8 total brand mentions per response in early May before falling to 2.6 within roughly six weeks. Its day-to-day variation during the study's stable period was also substantially higher than ChatGPT's.

A one-off check can therefore be misleading.

And tracking one engine tells you about that engine, not AI search as a whole.

The sources differ between engines too

The same pattern appears in citation data.

Vercite analysed 5.31 million citations collected between March and June 2026 from ChatGPT, Perplexity, Gemini, Google AI Overviews and Google AI Mode.

Among the 253 domains appearing in at least one engine's top 100 most-cited domains, only 23 appeared in all five.

That's 9%.

The most-cited platform also differed:

• >ChatGPT: Reddit
• >Perplexity: YouTube
• >Google AI Overviews: Google
• >Gemini: Shopify
• >Google AI Mode: YouTube

The dataset covered 158,847 unique domains.

This is why the relationship between mentions and citations becomes more useful when analysed by engine.

You are not trying to find one universal list of websites that determines AI visibility. Different systems pull from different parts of the web.

How to track AI brand mentions

Typing your category into ChatGPT once is useful for curiosity. It isn't reliable monitoring.

A better approach starts with a fixed set of prompts and repeats them over time – we walk through that process step by step here.

1. Choose the prompts that matter

Start with questions a prospective customer might genuinely ask.

For a CRM company, that might include:

• >best CRM for a small sales team
• >CRM for a growing B2B company
• >affordable CRM with automation
• >alternatives to Salesforce
• >easiest CRM to implement

Avoid building the entire dataset around prompts containing your own brand name. If you explicitly ask an engine about your company, appearing in the answer tells you very little about organic category visibility.

We've covered how to find and prioritize these questions in more depth in how to choose prompts to track.

2. Organize them with prompt tags

Different prompts represent different needs.

A CRM company might use tags such as:

Small business, Sales automation, Enterprise CRM, Alternatives, Integrations

You can then see that your overall mention rate looks healthy while discovering that you are nearly absent from one commercially important topic.

3. Track several AI engines

Run equivalent prompts across the engines your audience uses.

Vercite currently tracks:

• >ChatGPT
• >Gemini
• >Perplexity
• >Google AI Overviews
• >Google AI Mode

The goal is not to average away the differences. The differences are part of the information.

4. Store the full response

Keep more than a Boolean mentioned = yes.

You need the answer to see:

• >where you appeared
• >who appeared before you
• >which competitors were included
• >whether you were recommended
• >whether the wording was positive, neutral or hesitant
• >whether your brand appeared multiple times

5. Capture the citations alongside it

Then inspect which URLs and domains surfaced with those answers.

This connects the visible answer with the source layer you may be able to investigate through content, SEO, digital PR or outreach.

6. Track it over time

AI answers change even when the prompt stays exactly the same.

A trend across repeated runs is more useful than a screenshot from Tuesday afternoon.

Vercite runs tracked prompts repeatedly and keeps the underlying responses so changes in mentions, position, framing and citations can be compared over time.

What should you measure alongside mentions?

Brand mentions are a useful base metric. They become much more informative with context.

For each prompt or prompt group, look at:

Mention frequency
How often does your brand appear across repeated responses?

Mention position
Does the brand tend to appear early or late in the answer?

Competitor presence
Which brands repeatedly appear when you don't?

Framing and sentiment
Are you recommended, neutrally listed, qualified or criticised?

Citation sources
Which sources appear alongside answers where you and your competitors are mentioned?

Engine differences
Are you consistently visible in ChatGPT but rarely mentioned by Gemini or Perplexity?

These answer different questions. Combining them gives you a much better picture than trying to turn every aspect of AI search into one number.

Mentions also create an attribution problem

AI discovery does not always produce an AI referral.

Someone might ask ChatGPT for the best accounting platforms, see your company in the answer, close ChatGPT and Google your brand five minutes later.

Analytics might record that as organic search.

Another person might type the domain directly.

Someone else may never visit at all but remember the brand when they encounter it later.

This makes AI-driven discovery difficult to measure using click attribution alone.

One low-tech addition is surprisingly useful: add AI assistant / ChatGPT as an option when asking customers:

How did you hear about us?

It will not provide exact attribution. But it can reveal a discovery path that referral analytics frequently misses.

AI mentions tell you what made the answer

The simplest interpretation of a brand mention is still the most useful:

Your brand made it into the response the user saw.

From there, the interesting questions begin.

Were you first or last? Recommended or merely listed? Did competitors appear more often? Which sources surfaced? Does the same thing happen in Gemini? Will the answer look the same next week?

That's why Vercite treats mentions as one layer of AI visibility rather than a metric to look at in isolation.

Track the mention.

Then inspect the answer around it.