The practical AI Visibility guide
Measure competitor share of voice in a defined set of AI answers.
Share of voice can show which brands dominate your sampled answers. It becomes useful only when the question set, brand set, and counting rule are explicit.
Put it into practice
Set the counting rule
Share of voice is a ratio within your sample.
For this guide, AI share of voice is your brand’s qualifying appearances divided by the total qualifying brand appearances in a fixed comparison set. Count a brand once per completed answer. Decide whether “qualifying” means any visible mention or an explicit recommendation. Keep citations in a separate metric.
This is one transparent method, not a universal industry standard. Fix the comparison set before the review: your brand and three direct alternatives, for example. A result can change if you add brands to the denominator, even when your own appearances stay the same. State that the measure describes sampled answers, not customer market share or all AI users.
Work through the arithmetic
Distinguish your share from your appearance rate.
| Brand | Answers containing a qualifying appearance | Share of appearances |
|---|---|---|
| ExampleCo | 4 | 4 ÷ 20 = 20% |
| Acme | 7 | 7 ÷ 20 = 35% |
| Globex | 6 | 6 ÷ 20 = 30% |
| Wayne | 3 | 3 ÷ 20 = 15% |
| Total for the fixed brand set | 20 appearances across 10 answers | 100% |
ExampleCo’s mention rate is 4 ÷ 10 = 40%. Its share of voice is 4 ÷ 20 = 20% under this counting rule. Both can be correct because an answer can include more than one brand. Mention rates across brands can therefore sum to more than 100%.
Deduplicate aliases such as a company name and its product name according to a rule agreed before counting. Exclude failed requests, and report how many answers completed. If every answer contains no qualifying brand from the set, the denominator is zero: report no qualifying appearances rather than inventing a percentage.
Keep cohorts comparable
Do not let a changing sample masquerade as progress.
Use the same core prompts, analysis mode, model set, language, and brand set across comparisons. If you add a model or prompt group, report it separately before incorporating it into a new baseline. Keep the raw per-answer observations so a teammate can reproduce the count.
A provider with ten tested models should not silently carry ten times the weight of a provider with one model if the report is intended to give providers equal weight. Choose whether observations, models, or providers receive equal weight, and state that choice. Repeated runs should follow the same rule for every brand.
OpenAI’s evaluation guidance recommends task-specific evaluation and checking automated metrics against human judgment. Here, that means reading a sample of counted answers to check whether a name is a genuine recommendation, a neutral reference, or a warning.
Look beyond the ratio
Use AEOchamp to inspect recommendation strength by intent.
Open User Intent in AEOchamp and choose a question closest to an actual buying decision. Review the Top Competitors comparison, then expand the question to inspect individual model results. The competitor scores are recommendation assessments from the Organic analysis. They are not percentages of a market or the share-of-voice ratio above.
Compare competitor evidence with your Organic and Mentioned scores. If your business scores strongly when introduced but is weak without its name, investigate why the model associates competing products with that workflow. If both target scores are weak, check product fit and the clarity of the relevant explanation before treating the issue as awareness.
User Intent
Every buyer question.
Your place in the answer.
See where your business comes up, where a competitor takes the lead, and where being known still isn’t enough to be suggested.
Organic visibility
38/ 100Your brand is withheld from the question.
Mentioned score
79/ 100The model is given your domain.
Visibility gap
+41pointsKnown when named. Less visible unprompted.
Your score beside the market leaders
Five pinned buyer questions. Select a question to explore its model results.
| User intent | Organic | Mentioned | Gap | Top competitors | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
14 | 78 | Acme91Globex86Initech81 |
||||||||||||||||||
|
Inside this intent · individual model results3 illustrative model results
| ||||||||||||||||||||
32 | 84 | Umbrella90Hooli85Stark82 |
||||||||||||||||||
|
Inside this intent · individual model results3 illustrative model results
| ||||||||||||||||||||
61 | 88 | Globex89Acme86Wayne80 |
||||||||||||||||||
|
Inside this intent · individual model results3 illustrative model results
| ||||||||||||||||||||
8 | 57 | Umbrella88Stark84Hooli79 |
||||||||||||||||||
|
Inside this intent · individual model results3 illustrative model results
| ||||||||||||||||||||
75 | 88 | Umbrella87Wayne83Initech78 |
||||||||||||||||||
|
Inside this intent · individual model results3 illustrative model results
| ||||||||||||||||||||
Start with what buyers need.
Five baseline intents turn your business profile into concrete questions your potential customers might ask.
Find the missed recommendation.
A score of 14 organic and 78 mentioned is a +64 gap: a strong fit when named, but a weak presence without the prompt.
Go beyond the average.
Expand an intent to compare individual model results and see which competitors appear beside each score.
Learn from the competing answer
Turn the comparison into a customer-facing improvement.
| Finding in the answer | Page task |
|---|---|
| A competitor is associated with clear approval records. | Explain what your product records, show a real record, and state its limits. |
| The model favors a workflow for guest reviewers. | Document guest access requirements and demonstrate the review path if your product supports it. |
| Different models place your business in a different category. | Reconcile the homepage, About page, and core use-case descriptions. |
| The competitor offers a capability you do not. | Explain your actual fit honestly; choose a question where your product has a relevant advantage. |
Keep the improvement tied to the original question. A generic competitor-comparison page will not fix an unclear workflow by itself. Use the prompt selection guide to narrow the situation and the score interpretation guide to decide what the evidence supports.
Your questions, answered
What is AI share of voice?
For the method in this guide, it is your brand’s qualifying answer appearances divided by all qualifying appearances for a fixed comparison set of brands. Count each brand at most once per completed answer. Other methods exist, so always publish the definition.
Is share of voice the same as mention rate?
No. Mention rate uses completed answers as its denominator. This share-of-voice method uses qualifying brand appearances. Multiple brands can appear in one answer.
Does AEOchamp automatically calculate share of voice?
AEOchamp shows competitor recommendation scores and model-level comparisons in User Intent. It does not currently provide the mention-based share-of-voice calculation described here. Use its competitor evidence to guide your diagnosis and keep any separate share-of-voice calculation clearly labeled.
From advice to a clear next step
Learn why competitors appear for your buyer questions.
Use User Intent to compare competing brands and individual model results. Turn that evidence into a specific use-case page improvement.
Start your free AEOchamp check