Start with the question your next client would ask
If you ask ChatGPT about your own firm, you have supplied the answer with a powerful clue: the business name. That can be a useful check of how your public information is represented. It answers a different question from whether someone who has never heard of you would encounter your firm.
For a mortgage business, compare these illustrative prompts:
- “What does Example Mortgages specialise in?”
- “Which mortgage broker understands limited company directors in Brighton?”
The first is branded. The second describes a client situation and a location. Neither example is a recorded AI result. Together, they show why an AI Visibility review needs to distinguish discovery from recognition.
Keep three questions separate
Did the answer name your business? Did it provide a captured citation to your business? Did anyone visit your site or contact you afterwards?
These are different observations. A citation can help document a source used in an answer, but it does not demonstrate an enquiry. An unlinked mention should not quietly become a captured citation in your scorecard.
Digital Ella’s current numerical baseline is citation-based, reported separately for each system. Captured answer text provides context. The measurement method explains the scope, including the use of one run per question per system and how unavailable responses are handled.
Build the question set before you inspect the answers
Start by listing the services you actually provide, the client situations you understand and your agreed main location. Separate the questions that name your firm from those that describe a need.
For each useful question, record what you are trying to learn. A question about limited company directors might explore specialist discovery. A question about comparing local brokers might explore consideration and trust. Keep the intent clear enough that a competitor comparison means something.
Then agree the same question set for the firms you will compare. Avoid selecting only the questions where you already know you appear, or treating a single favourable answer as evidence of consistent visibility.
Record the limits alongside the result
AI answers can change with wording, date, location and system. Some questions do not produce a Google AI Overview. Some responses cannot be captured reliably.
Keep those gaps visible. An unavailable answer is not proof that your firm was absent. A single run provides a dated observation; it cannot tell you how often your business would appear across many independent runs.
Google’s guidance for generative AI search continues to emphasise useful content and SEO foundations. There is no special file or schema field that guarantees a recommendation.
Check what the prospective client can verify next
Suppose someone does encounter your business and follows up. Can they understand your services, find the person behind the firm and see how the next conversation works?
That is where Digital Trust — clear information and credible evidence becomes relevant. Visibility and confidence are connected parts of the journey. An unclear service page or conflicting contact details may deserve attention even when your business already appears in a captured answer.
Choose the next step from the evidence
A first check may uncover a factual error worth correcting, a specialist page worth clarifying or a need for a more structured baseline. It may also show that your public information is already clear in the questions you tested.
The useful outcome is a decision about what deserves work. Foundation and Growth audits provide defined levels of investigation and a review with Ella. You can take the priorities to your own team or discuss separately scoped implementation afterwards.
If you do a quick check yourself, keep the question, date, system, answer and sources together. Label it as an observation. That small habit makes the next conversation much more useful.