Definition
At Transform& we use AI visibility to mean one thing: across a defined set of buyer questions, measured in a named set of AI systems, how often does your company appear in the answer? Accuracy is tracked as a separate measure, because appearing and being described correctly are different problems.
Appearing can mean three different things, and they are worth separating. You can be mentioned as one of several names. You can be cited, where your own content is used as the evidence behind the answer. Or you can be recommended, where the answer actively suggests you as an option to consider.
These are Transform& definitions, not industry standards. There is no agreed measurement authority for AI answers, so anyone claiming a standardised score is describing their own method. We describe ours openly instead.
Why it matters commercially
Buyers increasingly arrive at the first sales conversation with a shortlist already formed. Part of that shortlist now comes from asking an assistant to compare options, explain a market or suggest who to talk to.
If your company is absent from those answers, you are not losing a ranking. You are being left out of the consideration set before anyone has evaluated you. That is a commercial problem long before it is a marketing problem.
How it works
AI systems do not all produce answers in the same way. Some retrieve current sources, some rely more heavily on model knowledge and some combine both.
A company can influence the quality of its public presence, but it cannot control whether a particular system includes it.
Crawler access is documented by the AI providers themselves. OpenAI — Bots and crawlers documentation
What can be measured
AI visibility can be observed through presence, citations, recommendations and accuracy across a defined scope.
The exact measurement design depends on the market, systems and engagement.
Common mistakes
- Testing brand-name questions only. If you ask about your own company, you will usually appear. That tells you almost nothing.
- Measuring once. Answers vary between runs; a single sample is an anecdote.
- Treating a mention as a win. Being named in passing is not the same as being recommended.
- Chasing volume. Publishing more pages does not make you more citable if the important claims stay unverifiable.
- Ignoring access. If a system cannot access the meaningful content on a page, that content cannot help your visibility.
What AI visibility cannot tell you
AI visibility tells you what is happening. It does not by itself tell you what needs to change.
That is where the wider Commercial Authority problem becomes important.
Relationship to search
AI visibility is not a replacement for search visibility. Much of the same public information supports both: accessible pages, clear claims, consistent facts and credible external evidence.
In practice, companies that were already unclear or unreadable to search engines tend to be invisible to AI systems as well. The work overlaps heavily; the measurement is what differs.
Transform& provides AI visibility analysis and improvement as part of its services, alongside search visibility and competitor analysis.
Limitations
Generated answers are non-deterministic, personalised in places, and change without notice as systems are updated. Measurement gives you a defensible picture over time, not a precise instrument reading.
We report ranges, sample sizes and dates. Where we cannot attribute a change to the work, we say so.
Sources and references
External documentation and research referenced for factual claims about how search engines and AI systems work. Definitions and measurement methods described as Transform& definitions are our own.