How we measure
A score of "visibility 37" is worthless unless you know how it was produced. So here is the method - what we ask, where we ask it, how many times, and how we turn the answers into a number you can compare over time.
We start from your customers' questions, not from keywords
Classic search runs on short queries. A conversation with an AI assistant is different - the questions are long, specific and full of conditions: budget, location, requirements, a comparison with an alternative.
So we do not start from a keyword list. We start from the real decisions your customers make. For every business we build a set of questions that covers:
- Finding a provider - "who offers X near me"
- Comparisons - "what is the difference between A and B"
- Checking a specific name - "is company X any good"
- A problem with no provider named - "what should I do if..."
- Conditions and constraints - budget, location, specific requirements
The set is built together with you and updated every month. It is the foundation of the whole measurement - if the questions are not the ones your customers really ask, nothing else matters.
Up to seven surfaces, because their answers differ
We cover ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Overviews and Google AI Mode. This is not about completeness - results differ significantly between them. A business can be well represented on one and completely absent from another, for different reasons.
How many of them are included depends on your tier: three in Basic, five in Extended and all seven in Full. Details on the packages page.
Every report breaks results down by surface. The combined number is useful for tracking the trend, but decisions are made from the breakdown.
What gets counted
Being mentioned is not the same as being recommended. So every answer is scored on four separate indicators.
Mention
Your business appears in the answer in any of its known forms. We also track spelling variants - legal name, trade name, abbreviations and common misspellings.
Citation
The answer refers to your website or to a source connected with you. A citation can exist without a mention and vice versa - so they are counted separately.
Recommendation
Your business is named as a choice, not merely mentioned. This is the indicator most directly tied to sales.
Share of voice
Of all the businesses named across the full question set, the share that is you. It is calculated against up to five competitors chosen with you, and it is the main number we track over time.
We also track how the model describes your business. An inaccurate or outdated description is a separate problem, fixed by different means.
How we deal with instability
AI assistants do not give the same answer every time. This is the main difficulty in measurement and we do not hide it. The same question can return different names on different days. So a single query is not a measurement - it is an anecdote.
Repetition
Every question is asked multiple times, not once. We record how often you appear, not a single result.
A stable set
The question set does not change at random between measurements. When new questions are added, the report flags them and the original ones are tracked separately, so the comparison stays valid.
Identical conditions
The same settings, language and context in every measurement.
Trend, not snapshot
One measurement shows a state. Three show a direction. We work with the direction.
Honest about the limits. Even so, differences of a few percent between two measurements should not be read as change. Our reports mark what is a significant difference and what is noise.
What happens after the measurement
The report shows where you stand, but it changes nothing on its own. For every question where you do not appear, we look for the reason. It usually sits in three places at once.
Technical foundation
How accessible your content is to the systems that read it, structure, structured data and speed. An important detail: most AI systems do not execute JavaScript and see only the initial HTML. Websites that look great in a browser and rank well on Google can be completely empty to them.
Content
Does your website answer the question the customer is asking - in a form that can be cited.
External signals
Everything the model uses to understand your business outside your own website: directories, review platforms, professional profiles, press and industry mentions. Consistency between them matters more than most businesses expect.
After the diagnosis we carry out the changes. We do not hand over a report full of recommendations - we work on the website, the content and the external sources.
What we do not promise
We do not promise a position. Nobody controls the answers these systems give. We work on the signals that influence them.
We do not promise a timeline. The first changes usually show within a few months, but that depends on where we start and how quickly changes are made to your website.
We do not promise direct traffic. AI influence often happens before the click - the buyer gets a short list, then searches your name or calls you directly. That rarely shows up in analytics as coming from an AI platform. So the main indicator is visibility, not traffic.
We do not measure what cannot be compared. If your market has too few comparable competitors, or one predetermined player will always be named, we will tell you and decline the work.
See where you stand
The check shows how often AI assistants mention your business for the questions your customers ask - and who they recommend instead.
Free · No commitment · Delivered within one business day