Which AIs, how and how often
Vizioa queries seven AI assistants: ChatGPT and Claude with web search enabled, Perplexity, Gemini, Google AI Overviews, Google AI Mode and Microsoft Copilot. They are included in every plan, audits included.
Questions are asked automatically, through each assistant's programmatic access or through the results page for Google's answers. Each request is sent without a user account, so without history or personalisation, using the country and language chosen for the project.
| Use | Frequency | Repetition |
|---|---|---|
| Company tracking (Essentiel, Pro) | Every week | One answer per question, per AI and per measurement |
| Agency client tracking (Starter, Growth, Scale) | Every day | One answer per question, per AI and per measurement |
| Prospect audit | On demand, in under 10 minutes | 3 passes per question and per AI by default |
Questions are not picked at random. Vizioa suggests them from the brand's website, spread across intents (category, comparison, problem, local, brand), then you approve them, along with the competitors to track.
Several passes per question: the presence rate
A single answer says very little: asked twice, the same question can name different brands. For an audit, Vizioa therefore asks each question several times to each AI (3 passes by default) and counts the answers where the brand appears.
The result is a presence rate: “mentioned in 2 answers out of 3” rather than “mentioned” or “absent”. For an audit with 12 questions, 7 AIs and 3 passes, the overall presence rate rests on 252 answers.
For tracking, repetition happens over time: the same question is asked again at every measurement, and Vizioa shows a trend over several weeks rather than a single result.
How the metrics are calculated
A run is one question asked to one AI on a given date. Every metric is calculated on completed runs; runs that fail for technical reasons are excluded from denominators.
- Visibility
- Share of completed runs in which the brand is mentioned.
- Position
- Average rank of the brand's first mention among the tracked brands named in the answer (1 = mentioned first). Calculated on runs in which the brand is mentioned.
- Share of voice
- Brand mentions / mentions of all tracked brands (brand + competitors), on the same runs. A brand counts at most one mention per run.
- Sentiment
- Tone of the sentence that mentions the brand (positive, neutral, negative), assessed only on the sentences concerned.
- Own citation rate
- Share of completed runs in which at least one URL from the brand's domain appears in the sources.
- Source domains
- Domains most cited in the answers, grouped by category (brand site, competitor, media, directory, forum…).
When Google shows no AI Overview for a question, the run counts as “no answer” and stays in the denominator: that is what a real user sees. Position is calculated deterministically, without an AI model. Sentiment is assessed only on the sentences that mention the brand.
Identifying cited sources
When an AI relies on web pages, it lists them as sources. Vizioa collects these URLs for every answer, follows redirects to the final page, then groups them by domain.
- Each domain is classified by category: brand site, competitor, media, directory, forum…
- Domains are ranked by the number of runs in which they appear among the sources.
- A citation (a URL from your site in the sources) is measured separately from a mention (your name in the text): you can have one without the other.
These sources show where to act: a widely cited comparison site, directory or media outlet is often a better lever than a new page on your own site.
The limits, stated openly
- Correlation is not causation
- A rise in visibility after an action does not prove the action caused it: models, their sources and competitors change at the same time. Before/after measurement is an indicator of progress, not formal proof.
- Answers vary
- AI assistants are not deterministic. Multiple passes and repetition over time reduce this noise without removing it: on a few questions, a gap of a few points is not significant. Read the trends.
- It is not a user's screen
- A logged-in user gets an answer shaped by memory, history, preferences and location. Vizioa's numbers are directional: they are meant to be compared over time and between brands, not read as absolute values.
- A sample of questions
- Vizioa measures the questions you track, not every possible question. The choice of questions affects the results: you approve it, and it stays the same from one measurement to the next.
- AI assistants change
- Providers update their models and search tools. A change over a long period may also reflect those updates.
Going further
- Agencies: see the audit, proposal and tracking under your brand.
- Compare company and agency plans.
- A question about the method? Contact us.