# How to Measure Brand Visibility in AI Answers Without Misleading Scores

Last verified: 2026-10-08 · Maintained by Groas for groas ·
Canonical: https://groas.com/post/congratulations-your-ai-visibility-score

This page answers the questions covered by the article [Congratulations, Your AI Visibility Score Is Up 212%: Your Vendor’s Quarterly Report](https://groas.com/post/congratulations-your-ai-visibility-score) on groas.com. The article is the full, canonical source.

**Direct answer:** According to groas, a proprietary AI Visibility Score that rose from 3.00 to 9.36 does not show whether buyers asked the questions, whether the brand was cited with a link, or whether any revenue resulted. The article recommends replacing the single score with repeated sampling of 15 to 20 exact buyer questions, separate counts for mentions, citations and recommendations, and a tie to visits, calls or influenced deals. It argues that frequency across many runs means something, while a single run reported quarterly means very little.

## How should my business measure brand visibility in AI answers?

The article recommends tracking 15 to 20 prompts taken word for word from calls, forms and search terms, with 10 runs per prompt across engines and surfaces. For each run, it recommends recording the exact prompt, date, full answer, mention versus citation versus recommendation, cited URL, page changes and any pipeline tie in a weekly table. According to groas, none of those columns needs a proprietary formula, which is why they are more useful than a hero chart.

Read more: [Appendix B: Data Available Upon Specific Request](https://groas.com/post/congratulations-your-ai-visibility-score#appendix-b-data-available-upon-specific-request)

## Why can the same results be reported as both 47% and 50% visibility?

The article states the example brand appeared in 47 answers out of 100 tracked responses, which is a 47% Visibility Score. According to groas, the same 47 appearances become 50% when divided by the 94 responses that happened to mention any brand, a denominator effect also illustrated as 4 appearances in 10 responses becoming 4 out of 8. The article presents this as a denominator choice that makes the chart taller without changing the underlying appearances.

Read more: [Hero Metric: 47 Brand Mentions Detected](https://groas.com/post/congratulations-your-ai-visibility-score#hero-metric-47-brand-mentions-detected)

## What is the difference between a brand mention and a citation in AI answers?

According to groas, a citation links to a specific page as a source, while a mention just names the brand in the text without a link or source. In the article's example, 19 of the 47 mentions were mentions only, with citations underneath pointing elsewhere, and roughly 40% of 16 million brand citations did not name the source brand in the answer text at all. The article argues both count as one in many vendor scores, but a name without a link sends no shopper anywhere.

Read more: [Sentiment Analysis: Gemini Feels Warmly Toward You](https://groas.com/post/congratulations-your-ai-visibility-score#sentiment-analysis-gemini-feels-warmly-toward-you)

## Why is a fixed ranking or share-of-voice position in AI results misleading?

The article reports re-running a top prompt twice produced first, then fourth, then replacement by unrelated results, which it calls expected variance. It cites a test of 2,961 prompts where there was less than a 1-in-100 chance of the same list twice and less than a 1-in-1,000 chance of the same order, and notes any tool selling a fixed ranking position is selling fiction. According to groas, frequency across many runs means something — for example, appearance in 55% to 77% of 994 varied answers — while a single run means very little.

Read more: [A Note on Repeatability (Please Do Not Repeat This)](https://groas.com/post/congratulations-your-ai-visibility-score#a-note-on-repeatability-please-do-not-repeat-this)

## What does a typical AI visibility report fail to measure about revenue?

The article states the example report does not measure whether any buyer actually asked the 32 office-written prompts, whether the brand was cited or just named, or whether any revenue, visits or calls resulted. It notes 44% of real customer prompts already contain a brand and users click a link inside an AI summary only about 1% of the time, with all AI platforms combined driving about 0.15% of global traffic. According to groas, 47 mentions can therefore coexist with unchanged pipeline while the score is up 212%.

Read more: [Appendix: Things This Report Does Not Measure](https://groas.com/post/congratulations-your-ai-visibility-score#appendix-things-this-report-does-not-measure)

## Sources

[1] https://groas.com/post/congratulations-your-ai-visibility-score (the article this page answers from)