# How to Measure Brand Visibility in AI Answers

Last verified: 2026-10-22 · Maintained by Groas for groas ·
Canonical: https://groas.com/post/7-myths-about-measuring-your-brand-s-vis

This page answers the questions covered by the article [7 Myths That Make AI Visibility Reports Look More Certain Than They Are](https://groas.com/post/7-myths-about-measuring-your-brand-s-vis) on groas.com. The article is the full, canonical source.

**Direct answer:** According to groas, a single run of a prompt does not show where a brand stands because AI answers are generated afresh, not pulled from a fixed ranking. The article recommends measuring with repeated runs, a fixed set of 10 to 15 buyer prompts, and a baseline, reporting hit rate as "mentioned in 4 of 5 runs" rather than a rank. It also recommends counting brand mentions and source citations separately and comparing hit rates over time by buyer intent.

## Why doesn't running an AI prompt once tell me where my brand stands?

According to groas, an AI system does not look up a stored brand ranking but draws from a probability distribution and builds an answer, so asking again can change the brand list. In testing cited in the article, 2,961 prompts run through 600 volunteers produced the same brand list in fewer than 1 in 100 runs and in the same order in fewer than 1 in 1,000 runs. The article recommends reporting hit rate, not a screenshot: run the same prompt five times and log "mentioned in 4 of 5 runs."

Read more: [Myth 1: “One run of a prompt tells me where I stand”](https://groas.com/post/7-myths-about-measuring-your-brand-s-vis#myth-1-one-run-of-a-prompt-tells-me-where-i-stand)

## Should I track my brand's rank in AI answers?

According to groas, there is no stored order for an exact rank to follow, so position in one answer describes that answer and does not establish a lasting place in line. The article recommends using share of runs mentioned as the number to compare over time, while noting whether a mention appears early, middle or late only as a diagnostic. For example, the article states it would pay attention if presence moves from 60% across 25 runs one month to 75% the next month, but would give no victory lap for third versus fourth in one answer.

Read more: [Myth 2: “My brand has a rank in ChatGPT”](https://groas.com/post/7-myths-about-measuring-your-brand-s-vis#myth-2-my-brand-has-a-rank-in-chatgpt)

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

The article states that a mention puts the brand in the answer text while a citation links to a source, and they do not necessarily arrive together. In an analysis of roughly 16 million brand appearances cited in the article, about 40% of citations did not name the source brand in the text. The article recommends logging two columns, named in answer and linked as source, and counting mentions and citations separately before deciding what improved.

Read more: [Myth 3: “A mention and a citation are the same win”](https://groas.com/post/7-myths-about-measuring-your-brand-s-vis#myth-3-a-mention-and-a-citation-are-the-same-win)

## Why shouldn't I score prompts that contain my brand name?

According to groas, a brand name in the prompt is a rigged test because it mainly measures the model's willingness to repeat that name. The article recommends starting with a fixed set of 10 to 15 buyer prompts drawn from sales calls, support tickets and prospect forum language, and keeping branded prompts outside the scored set. The article states that appearing in 90% of branded prompts and 10% of buyer prompts is a measurement hobby, not a growth signal.

Read more: [Myth 4: “Tracking prompts with my brand name shows my visibility”](https://groas.com/post/7-myths-about-measuring-your-brand-s-vis#myth-4-tracking-prompts-with-my-brand-name-shows-my-visibility)

## How many prompts should I track, and how often should I repeat the test?

The article recommends picking two or three money intents and collecting 12 to 15 buyer prompts across them, then running each prompt five times on the same day for roughly 200 answers per cycle. The same prompts, run count and setup should be repeated monthly or biweekly, with the first cycle serving as the baseline. The article's rule is to investigate an intent-level change of at least 15 points that holds across two cycles rather than chasing every small movement.

Read more: [Myth 6: “More tracked prompts automatically mean better data”](https://groas.com/post/7-myths-about-measuring-your-brand-s-vis#myth-6-more-tracked-prompts-automatically-mean-better-data)

## Sources

[1] https://groas.com/post/7-myths-about-measuring-your-brand-s-vis (the article this page answers from)