Blog
Brands

How to measure AI brand visibility: metrics, data quality and decisions

Separate mentions, recommendations and citations. Use example data to calculate a mention rate, check response quality and choose a practical improvement.

Illustration of sources used in AI answers

Updated on 13 October 2026.

Start with the decision

A useful report explains which customer need deserves attention and what evidence supports a change. Define the country, language, system, exact questions and observation window. Separate questions about a category from questions that name a brand. They test different situations.

Keep the same scope when comparing reports. If questions, models or collection methods change, record the change before interpreting a higher percentage.

Separate three kinds of visibility

A mention names the brand. A recommendation presents it as suitable for a stated need. A citation links to a source. A linked source may belong to another publisher, and a name can appear without an endorsement.

For a mention rate, divide usable responses with a confirmed mention by all usable responses in the same scope. Read the context and check sources separately. An undocumented composite score may use a different formula.

Example data: how do you calculate a mention rate?

Let’s work through a bicycle repair shop example. A list contains 50 records: 40 usable responses, 3 blank texts and 7 unavailable full texts. Eight usable responses mention the shop. These example data explain the method and do not represent any brand’s results.

The mention rate is 8 ÷ 40 × 100 = 20%. Why divide by 40 rather than 50? Only usable responses enter the calculation. Blank and unavailable texts stay outside the denominator because their outcomes are unknown. The rate describes this example’s scope, not all AI conversations.

Example dataValueMeaning
Listed records50All records in the example
Usable responses40Denominator
Blank texts3Excluded
Unavailable full texts7Outcome unknown
Responses with a mention8Numerator
Mention rate20%8 ÷ 40 × 100

Read search performance separately

Google Search Console reports impressions, clicks and positions for a defined search scope. Bing has its own reporting. These units are different from AI responses: do not add search impressions to mentions or infer visits from a citation.

A second example looks at search performance: 320 impressions and 16 clicks give a 5% CTR. This is a separate metric. It does not explain the shop’s AI mention rate. Check queries, country, device and period before changing a title. A missing export row does not automatically mean zero.

Google includes its AI search features in Search Console’s overall Web reporting. That reporting is not a count of ChatGPT recommendations. Website visits and completed business actions require their own analytics and attribution evidence.

Turn an observation into a task

Suppose an answer highlights another workshop’s weekend repairs. First check whether the shop in our example offers that service. If it does, clarify the hours on its existing service page. If it does not, leave that offer out.

Check whether AI sees your business

Run a free audit

The proposed task should name a page, a verified information gap, an owner and a review criterion. Check existing content and drafts before creating another article. A mention is a reason to investigate, not proof of why the system selected a source.

Make the report verifiable

Include question scope, dates, usable responses, blanks, unavailable outcomes and the definitions of each metric. Separate mentions, recommendations, citations, visits and commercial results. There is no universal visibility percentage that guarantees sales.

Use a separate before-and-after protocol to evaluate an update. One answer or a short increase cannot establish causation. Publish useful, checkable explanations rather than repeated pages or unsupported product claims.

For evaluating a change, read the before-and-after guide.

Google documentation: AI features

Google: creating helpful content

Share:

Read other articles about AI
eCommerce

How to integrate a store with AI without code?

Answer engines - ChatGPT, Google AI Overviews, Perplexity - are becoming a viable source of traffic and sales. Instead of hunting only for clicks, it's worth ensuring that your product sheets and purchase policies are understandable to models and readily cited by them. The good news: this can be done without a programmer, in minutes, using no-code tools (e.g. Semly.ai).

AEO

AI Engine Optimization (AEO) - the new SEO for e-commerce

In 2025, the rapid growth of AI in online commerce makes it a key tool not only in streamlining processes, but especially in generating sales and capturing customers from new channels. Traditional SEO, while still important, is losing effectiveness in the face of new search algorithms based on artificial intelligence (e.g., Google AI Overviews, ChatGPT, Gemini).

GEO

How do you get ChatGPT and other AI models to recommend your services?

More and more buying decisions today start not with Google, but with a question to ChatGPT, Gemini or Perplexity. This is where users ask for recommendations of specific service companies - law firms, agencies, software houses or medical specialists. If your brand doesn't appear in the models' responses, you're losing real leads. Here's a complete step-by-step GEO strategy to help you change that.

eCommerce

How do you get AI talking about your store?

Learn what to do to make AI models (like ChatGPT or Gemini) recommend your store and products. Learn key content structuring strategies and how to use the Semly platform to get free traffic that converts up to twice as well! We bring you a guide to Generative Engine Optimization (GEO).

Check if AI ChatGPT sees your brand

Get your first AI visibility report in minutes.