Research
How to Measure AI Share of Voice
A practical framework for measuring brand visibility across ChatGPT and AI search using prompts, mentions, citations, competitors, and commercial intent.
Businesses are beginning to ask:
“Are we showing up in ChatGPT?”
That question is useful, but incomplete.
A brand could appear for an obscure informational question and still be invisible during the conversations that actually lead to customers.
The better question is:
How much visibility does our brand have across the AI conversations that matter commercially?
That is what I mean by AI Share of Voice.
Why Traditional Rank Tracking Breaks Down
Traditional SEO gives us a familiar metric:
Position #4 for keyword X.
AI responses are more variable.
The answer may change based on:
prompt wording
geography
conversation history
current information
personalization
model behavior
available sources
That does not mean AI visibility cannot be measured.
It means we need to measure it differently.
Instead of treating one prompt as an absolute ranking, measure a standardized sample of prompts over time.
Step 1: Build the Prompt Universe
Start by identifying the questions prospective customers ask during their buying journey.
Suppose you are a cybersecurity company.
Your universe might include:
Educational
What is managed detection and response?
Problem-aware
Our internal IT team cannot monitor threats 24/7. What should we do?
Comparison
MDR vs MSSP for a 200-person company
Recommendation
Best cybersecurity firms for mid-market healthcare companies
Transactional
Who can implement managed detection and response for us?
These prompts do not have equal commercial value.
That leads to the next step.
Step 2: Weight Prompts by Intent
A generic educational prompt should not carry the same weight as:
“Which vendor should I buy?”
A simple scoring model could be:
1 — Informational
2 — Problem-aware
3 — Solution-aware
4 — Comparison
5 — Recommendation / Transactional
That creates a Commercial AI Share of Voice rather than a vanity metric.
Step 3: Establish the Baseline
Before changing anything, run the full standardized prompt set.
Record:
whether your brand appears
whether competitors appear
whether your website is cited
what external sources are cited
where in the answer you appear
whether the mention is positive, neutral, or negative
This becomes Day 0.
Without a baseline, you cannot credibly claim improvement.
Step 4: Calculate Mention Rate
The simplest metric is:
Brand Mention Rate
If you appear in 14 of 100 tracked prompts:
Mention Rate = 14%
Useful.
But still incomplete.
You may appear mostly in low-value prompts.
Step 5: Calculate Commercial Share of Voice
Apply the intent weighting.
Suppose:
Brand A appears frequently in educational prompts.
Brand B appears less often overall but dominates recommendation and comparison prompts.
Brand B may have significantly more commercially valuable visibility.
Your scoring system should reflect that.
A simple weighted score can assign each appearance the intent score of the prompt.
Then compare your total weighted appearances with the total appearances of the brands in your competitive set.
The exact formula matters less than using the same methodology consistently over time.
Step 6: Track Citation Rate
There is an important difference between:
ChatGPT mentions your brand.
and:
ChatGPT cites your website as a source.
Both matter.
Track them separately.
Mention Rate
How frequently the brand appears.
Citation Rate
How frequently owned content is cited.
The first measures brand visibility.
The second measures source authority.
Step 7: Measure Source Coverage
Create a ranked list of external domains frequently cited across your prompt universe.
Then measure:
how many mention you
how many mention competitors
which high-frequency sources you are missing
This turns AI tracking into an authority-building strategy.
Step 8: Track Referral Traffic
Visibility is interesting.
Traffic is better.
Leads are better still.
Revenue is the goal.
Track AI referral traffic separately in analytics.
For ChatGPT Search, OpenAI currently includes utm_source=chatgpt.com in referral URLs, helping publishers identify inbound traffic.
Then connect those visitors to:
form submissions
booked calls
trials
purchases
pipeline
revenue
Eventually the question should become:
How much customer acquisition is being influenced by AI?
Not merely:
“How often did ChatGPT say our name?”
Step 9: Re-Test on a Schedule
I like:
Day 0
Day 30
Day 60
Day 90
for an initial experiment.
For ongoing client programs, weekly or monthly sampling may make sense depending on the size of the prompt universe.
The important thing is consistency.
Use the same core prompts.
If you constantly change the test set, comparisons become meaningless.
A Simple AI Visibility Dashboard
A practical dashboard could contain:
AI Mention Rate
18%
AI Citation Rate
6%
Commercial AI Share of Voice
22%
Competitor A
31%
Competitor B
17%
High-Value Sources Covered
9 / 24
ChatGPT Referral Sessions
214
AI-Attributed Leads
12
Now AI visibility becomes something executives can understand.
Measurement Is the Beginning, Not the End
The purpose of tracking is not to produce colorful charts.
It is to identify actions.
For example:
Competitor A dominates “best provider” queries because it appears across five publications that ChatGPT repeatedly cites.
That creates a strategy.
Or:
Our website is cited frequently for educational content but we rarely appear in recommendation prompts.
That creates a different strategy.
Measurement should tell you what to do next.
The Bigger Opportunity
AI discovery is still early enough that the measurement standards are not settled.
That creates an opportunity.
Businesses that begin building baselines now will have something many competitors will not:
historical data.
They will know:
where they started
what actions were taken
what changed
which sources mattered
which content earned citations
whether AI visibility produced customers
That information may eventually become significantly more valuable than a single ranking.