AI visibility measurement records whether an organization appears in generated answers, which page or outside source supports the answer, what the answer says about the organization, and which competitors appear instead. For a medical organization, those observations should be separated between Google AI Overviews and assistants such as ChatGPT and Gemini because the discovery journeys and inputs are different.
What should an AI visibility baseline record?
A useful baseline preserves the answer, its context and the conditions under which it appeared. A simple yes-or-no mention count is not enough to explain what should change.
For every tested question or prompt, record:
- the platform and visible product or model label;
- the market and language used for the test;
- the exact query or prompt;
- whether the clinic, a clinician, a service or a location appears;
- whether the answer links to the clinic or cites another source;
- which competitors appear;
- what the answer says about the clinic, including neutral, favorable or unfavorable language;
- the date of the observation.
The date is part of the evidence, not a promise that the same answer will remain visible. Generated answers can change as platforms, sources and the wording of a question change.
Why should Google AI Overviews be measured separately?
Google AI Overviews belong to a search journey. The starting unit is a search query, and the generated answer appears alongside other Google results. Measurement therefore needs the query, the conventional search result, the cited pages and the clinic’s appearance in the overview.
This is Door 2 in Rotgar’s model. It is related to traditional SEO, but it is not measured only through rankings. The useful AI Overview signals are presence, citation, source selection, competitor visibility and sentiment. See the dedicated approach to Google AI Overviews.
How is ChatGPT and Gemini visibility measured?
ChatGPT and Gemini are conversational journeys. The starting unit is a prompt rather than a conventional search query, and a patient may move from an initial research question to a narrower provider-selection question through follow-ups.
This is Door 3, delivered through ChatGPT + Gemini Visibility. A prompt set should cover the stages that matter to the organization, such as understanding a treatment category, comparing approaches, identifying the appropriate type of specialist and finding a provider in a market. The set is tailored to the clinic rather than capped at an arbitrary number.
For each prompt, record brand presence, citations or links where shown, named competitors and the language used about the clinic. Our guide to AI visibility explains the difference between a mention, a citation and a recommendation.
How should prompts and queries be selected?
Selection should reflect the clinic’s services, patient decision stages, markets and languages. A large list of loosely related prompts creates activity but not a useful baseline.
Start with four inputs:
- Priority treatments or clinical services.
- The market and locations in which patients choose providers.
- The language patients actually use, not a literal translation of an English list.
- The decisions the clinic wants to understand: early research, comparison, specialist selection or local provider selection.
Public information is sufficient to start. Read-only Google Search Console access can be added later, with the clinic’s approval, when more precise clinic-specific Google Search data would improve the analysis.
Which metrics are useful?
Use transparent counts that can be traced back to captured answers. A practical measurement set includes:
| Signal | What it answers |
|---|---|
| Presence rate | In what share of the tested questions is the clinic named? |
| Citation rate | In what share is the clinic’s site used or linked as a source? |
| Recommendation rate | When does the answer actively include the clinic as an option? |
| Competitor gap | Which questions name competitors but omit the clinic? |
| Source gap | Which outside sources shape the answer, and where is the clinic absent? |
| Sentiment | How is the clinic described when it appears? |
Keep mention, citation and recommendation separate. Combining them into one proprietary score can hide whether visibility is genuinely useful. If a summary number is needed, document its formula and retain the underlying observations. The AI visibility score guide shows a transparent approach.
How do you make repeated checks comparable?
Rechecks are meaningful only when their conditions are sufficiently consistent. Preserve the wording, market, language, platform and classification rules. If a platform or model label changes, record the change instead of treating the new result as directly identical.
Before setting priorities, repeat any high-impact or unstable finding. A single surprising answer is a reason to investigate, not enough evidence for a strategy. Over time, the decision should be based on patterns across the relevant question set and the source pages behind those patterns.
What AI visibility measurement cannot prove
AI visibility does not by itself prove traffic, qualified enquiries, appointments or revenue. Those are later business outcomes. When a clinic can provide aggregated conversion information, it can help evaluate whether visibility work is attracting more relevant demand, but the AI baseline remains a measurement of discovery and representation.
It also cannot guarantee future citations. Rotgar measures observable results and prioritizes the signals a clinic can improve; no agency controls a platform’s generated answer.
Start with a free baseline
The free Google + AI visibility audit captures the clinic’s starting point across Google Search + Maps, Google AI Overviews, ChatGPT and Gemini.
Get a free audit or review how to improve AI visibility after the baseline is established.