How Often Does ChatGPT Recommend Your Brand? How to Measure It

To measure how often ChatGPT recommends your brand, you need more than one question and a screenshot. Define a stable set of prompts that represents real buying decisions, run each prompt several times under comparable conditions, and count how many valid answers include your brand.

Short answer: divide the number of valid answers that mention your brand by the total number of answers analyzed, then multiply by 100. That is your mention rate. If you separately count answers that clearly recommend the brand, you get a recommendation rate that is more useful for business decisions.

A manual check tells you whether you exist. Repeatable measurement tells you how often you appear, whether the trend is improving, and which competitors take your place.

What "how often" actually means

Frequency is not the raw number of times you see your name. It needs a denominator. Saying "ChatGPT mentioned us seven times" means little unless you know whether you analyzed ten answers or two hundred.

Use four measures:

• Mention rate: answers that include your brand divided by all valid answers.
• Recommendation rate: answers that present your brand as a suitable option divided by all valid answers.
• Prompt coverage: distinct prompts where you appear at least once divided by all prompts.
• Stability: the percentage of repeated runs of the same prompt where the brand appears again.

Example: you run 10 prompts three times and collect 30 answers. Your brand appears in 12 but is clearly recommended in only 8. Mention rate is 40%, recommendation rate is 26.7%, and if you appear in 6 of the 10 prompts, prompt coverage is 60%.

There is no universal percentage that is "good" for every category. A niche brand, an ecommerce store, and a bank compete in different markets. Compare the same prompt set over time and benchmark the result against competitors.

Three ways to measure it

1. A manual test for an initial signal

Create 10 to 20 high-intent questions, run each at least three times, and record the answers in a spreadsheet. This is a useful way to establish a small baseline without buying a tool.

The limitation is material: it takes time, comparable conditions are hard to maintain, and it does not build history automatically. One conversation should never be treated as representative of every user.

2. A spreadsheet or API workflow

You can structure prompts, answers, and dates in a spreadsheet or automate requests through an API when its terms allow it. This improves consistency and makes rate calculations easier, but you still need to normalize brand names, distinguish true recommendations, record competitors, and review ambiguous answers.

This option suits a technical team that wants to own the methodology and accepts the maintenance work.

3. A recurring AI visibility tracker

A tracker automates the prompt set, stores every answer, detects mentions and position, compares competitors, and shows changes between periods. It is the right option when you need recurring measurement across categories, markets, or brands.

Before choosing one, read our guide to evaluating an AI visibility tracker. This page does not compare vendors; it explains the method any solution should follow.

Step-by-step method

Step 1. Define prompts that represent real decisions

Do not start with your brand name. Asking "What do you think of Brand X?" forces the model to discuss it and does not measure discovery. Use prompts a buyer could ask without knowing you:

• "Which tools do you recommend for [specific problem]?"
• "What are the best [category] options for [customer type]?"
• "Compare [service] providers in [market]."
• "Which company would you choose for [use case] with [constraint]?"

Group prompts by intent: discovery, comparison, purchase, and specific problem. This shows not only whether you appear, but at which stage ChatGPT recommends you.

Step 2. Fix the measurement conditions

Record the date, language, market, model, and exact prompt text. Do not mix English and Spanish prompts in one rate. Do not compare two periods after changing half of the prompt bank.

ChatGPT evolves and not every variable can be controlled. The goal is not to eliminate variation; it is to document conditions well enough for a fair comparison.

Step 3. Repeat every prompt

A single answer may change when you run the prompt again. Running each prompt three times provides an operational baseline for distinguishing stable visibility from an accidental mention. Increase repetitions and the time window for higher-stakes decisions.

Do not run one prompt dozens of times while ignoring the rest of the customer journey. Covering multiple intents with a consistent protocol is more useful.

Step 4. Record more than yes or no

For every answer, store:

• Whether the brand appears.
• Clear recommendation, neutral mention, or negative comment.
• Position in the list or text.
• Competitors mentioned before and after it.
• The reason ChatGPT gives for recommending each option.
• Incorrect or outdated information.

Position needs its own interpretation. Read why position in AI answers matters. If ChatGPT always mentions you last, frequency alone may look more positive than the real outcome.

Step 5. Calculate and compare

Calculate rates for the full period and for each intent cluster. Compare the brand with competitors and with its own baseline. To quantify relative presence, use the AI Share of Voice framework.

A minimum monthly table should include total answers, mentions, clear recommendations, mention rate, recommendation rate, average position, and the leading competitor.

Why ChatGPT changes its answer

Generative answers are not fixed rankings. They may vary with prompt wording, conversation context, model updates, language, market, and newly available information.

Variation does not make measurement useless. It means measurement needs samples and trends. Nobody evaluates conversion performance from one visit; a brand should not evaluate ChatGPT visibility from one answer either.

Keep a stable prompt bank and repeat it weekly or every two weeks. During a launch, rebrand, or reputation issue, more frequent checks may be appropriate.

How to interpret the result

• High and stable frequency: ChatGPT consistently recognizes the brand for those intents. Review position and reasoning to see whether the recommendation is competitive.
• High but unstable frequency: visibility depends too much on wording or timing. Strengthen signals that explain your category and value proposition.
• Low frequency with one dominant competitor: analyze the sources, proof, reviews, and content supporting that competitor. Use a competitive audit in ChatGPT and Perplexity.
• Mentions with wrong information: fix sources of truth first. See how to correct wrong AI information about your company.
• No mentions: confirm the prompts are relevant, audit the entity, and establish a baseline before making isolated changes.

How Mentio measures it

Mentio lets you define a brand, competitors, and the questions that matter in its market. It runs recurring checks, keeps the answers, and summarizes mentions, position, framing, and change over time.

The value is not a decorative score. It is being able to say: "We appear in 38% of purchase-intent answers, up from 22% last month, while Competitor A still leads price-related prompts."

You can start with a visibility audit and use the result as your baseline. For wider measurement across ChatGPT, Gemini, and Perplexity, continue with how to analyze brand visibility across multiple AI models.

Frequently asked questions

How many prompts do I need to measure my brand in ChatGPT

For an initial baseline, use 10 to 20 high-intent prompts and repeat each at least three times. Expand the sample when you cover multiple categories, markets, or customer types.

How often should I repeat the measurement

A weekly or every-two-weeks cadence can reveal changes without reacting to every isolated answer. During launches or reputation incidents, more frequent measurement may be useful.

Does one ChatGPT query count as measurement

It counts as a point-in-time check, not a reliable frequency measurement. Answers can vary, so you need multiple prompts, repeated runs, and a known denominator.

Are mention rate and recommendation rate the same

No. Mention rate counts any valid appearance of the brand. Recommendation rate counts only answers where ChatGPT presents the brand as a suitable option for the stated need.

Can I rank well in Google and still be absent from ChatGPT

Yes. Google and ChatGPT do not order brands in the same way. Web authority helps, but entity clarity, external mentions, reviews, and information consistency also matter.

Can Mentio measure frequency automatically

Yes. Mentio repeats a prompt bank, stores the answers, and calculates your brand's presence against competitors so that periods can be compared with the same method.

Build a baseline you can repeat

The useful question is not "Did ChatGPT mention us today?" It is "How often does it recommend us in prompts that influence a purchase, and how is that percentage changing?"

Analyze my brand visibility

Want to know if AI mentions your brand?

Discover your visibility in ChatGPT, Claude and Gemini in minutes.

Analyze your brand for free

Related articles