How to measure your brand’s AI visibility

Build a baseline around buyer questions

Start with a decision your customer needs to make. For a scheduling product, an unbranded question could be “Which scheduling tools work for a small consulting team?” A branded question such as “Does Acme Scheduler support team booking?” answers a different question about accuracy. Keep these groups separate when interpreting results.

1. Choose a small, stable prompt set. Include category discovery, use-case evaluation, direct comparisons, and questions about your own product. Use customer language from sales or support rather than assuming a generated prompt has search volume.
2. Record the conditions. Keep prompt wording, language, intended market, selected providers, competitor set, and reporting dates with your baseline. Record changes before comparing periods.
3. Inspect the responses. Check brand matches, recommendations, ordering, and source links. A name collision or a critical mention can make a headline score misleading.
4. Compare like with like. Use equal reporting windows and the same provider mix. Note missing responses, failures, and cached responses; repeated checks are not automatically independent observations.

Understand what each metric counts

In MyMentions, one stored response from one provider is an observation. A prompt checked across three providers can contribute three observations. These definitions explain the GEO metrics; always inspect the reporting window and underlying responses.

Worked example: hypothetical data

Suppose 10 collected responses mention your brand 4 times and contain 6 recorded competitor mentions. Your mention rate is 4 ÷ 10 = 40%. Your share of voice is 4 ÷ (4 + 6) = 40%. If 2 responses recommend you, the recommendation rate is 20%. These equal mention-rate and share-of-voice values are coincidental: their denominators measure different things.

If the next period has only 5 responses because collection failed, compare coverage before interpreting a score change. Zero collected responses means no evidence; a displayed zero is not evidence that AI never mentions the brand.

Provider coverage, plans, and limits

Starter includes OpenAI, Google Gemini, and Perplexity. Pro and Enterprise also include Anthropic Claude, xAI Grok, Azure OpenAI, DeepSeek, and Meta Llama through Together.ai. MyMentions collects responses through provider APIs. Models, system instructions, retrieval, personalization, and timing may differ from a person’s experience in a consumer app.

The Google integration is Gemini, not Google AI Overviews or AI Mode. The Azure OpenAI integration is not a measurement of Microsoft Copilot. Coverage of a provider does not imply coverage of every search or chat experience carrying that company’s name. Source links are available only when the returned response provides them.

One provider check is one prompt run against one provider: 5 prompts × 3 providers = 15 checks. Plans include a 7-day free trial. See the full plan features and subscription terms before choosing. Listed prices and coverage were reviewed on October 27, 2026.

For referral reporting, install MyMentions’ tracking script on your website. It classifies detectable source signals and records visits and engagement. Missing referrers, script blocking, and classification errors can limit the data. Use your own conversion records to assess business outcomes.

Evaluate a tool against your actual monitoring needs

Compare the unit being sold, the collection method, and the exact surfaces covered before comparing headline prices. A tracked prompt, a provider check, and an AI answer allowance are not necessarily equivalent. Ask whether you can inspect raw responses, separate branded prompts, and preserve a consistent baseline.

Disclosure: MyMentions publishes this guide and sells one of these products. Competitor observations are from the linked first-party pages, checked October 27, 2026; they may change. This is an evaluation checklist, not a hands-on benchmark, exhaustive feature comparison, or ranking.

Turn observations into useful work

When an answer describes your product incorrectly, check whether your public feature, pricing, and support pages provide a clear answer. When competitors are cited, inspect the cited pages and ask which customer question they answer well. A citation does not establish why a model chose it.

Google’s guidance says its AI features use the established Search fundamentals: accessible pages, useful textual content, internal links, and structured data that matches the page. It does not require special AI markup or an AI text file. Eligibility does not guarantee inclusion. Read Google’s AI features guidance.

• Choose one clear change, such as explaining a missing integration or clarifying a plan limit.
• Record what changed and when, then collect a comparable later sample.
• Inspect the answers as well as the score. Other changes in models, sources, or competitors can affect the result.
• Review qualified signups or purchases separately from visibility and referral visits.

Sources and editorial policy

MyMentions maintains this product guide. Product definitions reflect the billing, provider coverage, and metric calculations reviewed on the date above. External product facts link to the publisher’s own documentation. Recommendations are editorial judgment; worked examples are hypothetical and contain no customer data.

We update the review date when a substantive review changes or reconfirms the content, rather than on every site build. Send a correction to support@mymentions.org. Explore more topics in the MyMentions blog.