AI Visibility Tracker: Methodology & Metric Reference
This article is a complete resource to learn what 3BL's AI Visibility Tracker measures, how AI platforms influence brand reputation, and the way we calculate every score.
What the AI Visibility Tracker Measures
As more people turn to AI assistants for research and recommendations, visibility within these experiences is a meaningful driver of brand recognition, reputation and trust. The AI Visibility Tracker from 3BL monitors how often your brand appears in responses generated by top AI platforms when they’re asked questions that influence reputation.
Reputation Drives Measurement
The ways someone rates a brand’s reputation are quietly built across a range of daily interactions — from news consumption and research, to firsthand experiences with the brand’s products or services, to the things that others say. AI models learn from and retrieve those same public signals when they answer questions, and what they say in response influences how the user sees the brand — creating a feedback loop that is quickly transforming how brands are perceived in ways many don’t yet realize.
Fortunately, public signals are something brands can influence through consistent storytelling.
3BL's AI Visibility Tracker measures how companies show up in AI-generated answers to the questions people ask that shape how they feel about a brand, from whether it’s seen as a good value, to how fast it innovates, to how it treats communities, workers and the environment. Measuring AI visibility through the lens of reputation connects the metric to the work comms teams already do, and surfaces opportunities to do it better.
How the framework is structured
The AI Visibility Tracker methodology and prompt library is built around seven broad criteria (called pillars) that are widely used in academia and industry to measure corporate reputation. These criteria extend to 29 thematic subtopics, grounded in 3BL’s 100 Best Corporate Citizens methodology and academic research, that reflect the range of questions people ask that ultimately shape their perception. See AI Visibility Tracker Definitions for the full list.
Every prompt the tracker runs is associated with a pillar and subtopic in the framework. Each subtopic is tested from the same set of angles to produce comparable results. Prompts for every subtopic in the framework cover:
- Which companies lead on the topic
- How companies in an industry compare
- Which companies AI recommends
- Examples of companies doing it well
- Critical or negative questions
- Questions that name the brand directly
Every prompt is also tagged by the audience most likely to ask it, such as investors, talent, consumers or media.
How prompts are scored
Standard prompts: The same core prompts run for every brand in an industry, so peer benchmarks are apples-to-apples.
Custom prompts: Run alongside the standard set to cover a brand's specific programs, products or priorities.
Branded prompts: Prompts that name the organization directly are excluded from core scoring so the scores reflect whether AI brings a brand up without being asked about it.
The AI Visibility Score
The AI Visibility Score is the primary metric in the tracker. It measures the share of AI responses that include your brand — even when your brand is not named in the question — across industry and general business queries tied to reputation. The score is shown as a percentage out of 100 to reflect the share of responses that mention your brand.
Other metrics found in the tracker — such as Named First, Share of Voice and Favorability — are measured separately and are not blended into the overall AI Visibility Score.
All scores are based on at least 30 AI responses. A score based on fewer than 30 responses displays "fewer than 30 answers" instead of a number in the corresponding field.
Each score is shown with a range (i.e., 21.9%-23.4%) which indicates how much the score could move by chance. Two companies whose ranges overlap are effectively tied.
Visibility Leadership Indicator
In addition to the numeric score, the AI Visibility Tracker includes a Leadership Indicator classification that represents where your score falls relative to others.
Leadership Indicator classification is determined relative to each company's peer set. Companies with "Leader" status have the highest scores among the set. Companies with scores that are at least half of the set's Leaders receive "Contender" status. "Trailing" status is assigned to companies with scores less than half of the Leaders in the set.
|
Topline score range |
Classification |
What it means |
|
Highest among each company's custom peer set |
Leader |
Strong, consistent visibility across tracked queries compared to the average |
|
At least half of the set's Leader scores |
Contender |
Meaningful presence in AI responses, with room to expand coverage and position |
|
Less than half of the set's Leader scores |
Trailing |
Appearing in some responses, but limited reach or prominence |
Supporting Metrics
Supporting metrics within the AI Visibility Tracker provide more information about how your brand appears in AI responses.
Named First
Shown as a percentage, the Named First metric shows how often your brand is mentioned first in an AI response, out of all the responses that mention you. For example, if a company is named in 1,115 answers, and named first in 761 of them, that’s equivalent to a 68% Named First rate.
Share of Voice
Shown as a percentage, Share of Voice reflects the rate at which your brand is mentioned across all tracked responses, relative to your named peer set, representing your share of the overall conversation relative to your competitors on topics of interest.
Favorability
Shown on a scale of 100 to -100, Favorability reflects the tone of AI-generated language when your brand is mentioned.
How Favorability is calculated
The Favorability Score is measured at the response level and centers around language used in association with the target brand, not with competitors, the industry, or problems in general. Favorability is not measured at the mention level. Every mention of the target brand in a single response will be tagged with the same Favorability Score.
The response-level scores are compiled in the tracker to produce a single Favorability Score that reflects how your brand is discussed most often on a given AI platform or in association with a given topic.
The score is calculated by subtracting the percentage of unfavorable responses from the percentage of favorable responses, out of all responses that name the company. For example, AI names a company in 100 answers: 45 are favorable, 25 unfavorable, 20 neutral and 10 mixed. Favorability = 45% − 25% = +20.
Importantly, the Favorability Score does not simply reflect positive or negative keywords. A response can mention problems without reflecting negative favorability or sentiment toward the brands addressing those problems. For example, a response that includes the phrase “climate change is devastating, and Apple is leading with net-zero pledges" results in positive favorability for Apple, not negative favorability simply because a problem (“climate change”) and a negative keyword (“devastating”) are mentioned.
The favorability model analyzes whether the brand's actions are favorable, not whether the topic itself is positive.
Positive (Scores 1 to 100): Your brand is portrayed favorably
Example: "Levi Strauss has reduced emissions by 40%."
Example: "Nike has committed to ending child labor in supply chains."
Negative (Scores -1 to -100): Your brand is portrayed unfavorablly
Example: "Target has faced multiple data breach lawsuits."
Example: "BP's oil spill in 2010 caused massive environmental damage."
Mixed (Score: 0) Your brand receives both praise and criticism, a likely scenario that follows the real-world nuance brands navigate each day .
Example: "Apple leads in accessibility features, but has been criticized for supply chain practices."
Neutral (Score: 0): Your brand is mentioned in purely factual terms without positive or negative association.
Example: "Walmart has 4,700 stores nationwide."
Example: Article discusses ESG trends without specifically praising or criticizing your company
Unknown (Score: 0): The AI couldn't determine favorability. This is rare and usually indicates a data quality issue.
|
Favorability label |
Score value |
Description |
|
Positive |
1 |
Favorable language is used in the response when your brand is mentioned |
|
Mixed |
0 |
Both positive and neutral/negative language is present |
|
Neutral / Unavailable |
0 |
No clear sentiment in the response or the sentiment is not evaluable |
|
Negative |
-1 |
Unfavorable language is used in the response when your company is mentioned |
How Prompts Work
Prompts are the questions the tracker asks AI models to see whether, and how, your brand shows up in the answers.
Prompt logic
Every prompt the tracker runs belongs to a subtopic, and every subtopic sits under one of seven reputation pillars. See AI Visibility Tracker Definitions for the full list of pillars and subtopics.
Each subtopic is tested from the same set of angles to produce comparable results:
|
Angle |
Example |
User Intent |
|---|---|---|
|
Who leads |
Which [industry] companies are leaders in business ethics? |
Asking who to trust |
|
How companies compare |
How do [industry] companies compare on employee wellbeing? |
Comparing companies |
|
Who AI recommends |
Which [industry] companies would you recommend for responsible sourcing? |
Asking what to choose |
|
Real-world examples |
What are examples of [industry] companies investing in their communities? |
Asking for examples |
|
Critical questions |
Which [industry] companies have faced criticism over data privacy? |
Asking who to trust |
|
Questions about you |
What is [your brand] known for on providing customer value? |
Asking who to trust |
Prompts are written with placeholders for your brand, industry and competitors, which the tracker fills in for your account. That's how the same standard prompt produces a relevant question for every company in an industry. Some prompts are simple, and some are more complicated or advanced, to reflect the ways real users across backgrounds ask AI questions.
Every prompt is also tagged in two ways:
Audience: The stakeholder group most likely to ask a given question, such as investors, talent, consumers or media. This effects language used in prompts based on the audiences you choose in your settings.
Framing: Whether the question is positive, neutral or critical. Critical questions show whether AI associates your brand with a topic's risks, not only its opportunities.
Types of prompts
|
Prompt Type |
Definition |
Counts Toward Visibility Score |
|---|---|---|
|
Standard |
Runs for every company in an industry, so results can be benchmarked against peers |
Yes |
|
Custom |
Added for a specific company's programs, products or priorities; runs alongside the standard set |
No; counts toward a separate score for those specific prompts |
|
Branded |
Names the company directly |
No |
|
Unbranded |
Asks about an industry or topic without naming the company |
Yes |
Branded prompts are excluded from core scoring so your score reflects whether AI brings up your brand without being asked about you directly. Branded results appear in a dedicated section and are not included in calculating other scores.
Adding custom prompts
The standard library covers most of what companies need. Add a custom prompt when there's a question it doesn't ask or is specific to your company.
To write a strong custom prompt:
- Write it like a real person would. Plain language, one question, no jargon.
- Tie it to a topic. Every custom prompt should fit one of the 29 subtopics in the framework so it rolls up correctly in your results.
- Leave your name out where you can. "List the top companies in [industry] based on reliability, pricing, and customer satisfaction." tells you more about visibility than "How does [your brand] rate on satisfying customers?"
- Keep the framing neutral unless you're deliberately testing a critical angle.
- Check it's not already covered. If a standard prompt asks the same thing, you don't need a custom one.
Talk to your CSM to learn more about adding custom prompts to your tracker dashboard.
Key Terms
|
Column |
Description |
|
Prompt |
The exact question asked to the AI provider |
|
User Intent |
The nature of the question asked to the AI provider, classified into four categories:
|
|
Prompt Subtopic |
Subject matter of the query (e.g., business ethics, employee wellbeing, environmental stewardship) |
|
Citations |
Source URLs or references included in the response |
|
Peer Group |
Set of competitor companies tracked, if set |
Full Metrics Reference
The following table lists all currently defined metrics in the AI Visibility Tracker.
|
Metric name |
Format |
Description |
|
Visibility Score |
% |
Share of AI responses that include the company, — even when your brand is not named in the question — across industry and general business queries tied to reputation |
|
Leadership Indicator |
Label |
Leader / Contender / Trailing classification relative to threshold |
|
Named First |
% |
% of tracked prompts where the company is named and is the first mentioned |
|
Share of Voice |
% |
Rate at which the company is mentioned across all tracked responses, relative to their named peer set |
|
Favorability |
-100 to 100 |
Average sentiment across all evaluated responses |
|
Answers |
Count |
Number of AI responses collected |
The following table lists supporting metrics within the dataset that are used to calculate other metrics but are not displayed directly.
|
Metric name |
Format |
Description |
|
Company Mentions |
Count |
Total target company mentions |
|
Competitor Mentions |
Count |
Total competitor mentions |
|
All Mentions |
Count |
Total mentions across all tracked companies |
|
Company Citations |
Count |
Number of times the company is cited as a source |
|
Competitor Citations |
Count |
Number of times competitors are cited as sources |
|
Total Citations |
Count |
Total citations across all tracked responses |
Important Caveats
This methodology is observational, not causal. The Visibility Score reflects how AI platforms represent a brand in their responses at the time of measurement. It does not predict or guarantee future outcomes.
The tracker only measures. It shows AI responses exactly as given. Generated responses may contain mistakes. 3BL does not verify the accuracy of model responses; it only shares the results.
Metric definitions, score weights, and thresholds are subject to change as the product develops. We will communicate any significant methodology changes through your account team and in product release notes.
Questions about this methodology or your results? Contact your 3BL Client Success Manager.