Every point has a reason.
We assess the information patients use to find and choose a clinic. These are discoverability measures, not assessments of clinical quality or predictions of Google’s algorithm.
Four separate views
Profile readiness, reputation, listings health, and AI visibility answer different questions. We keep their scores separate. All capabilities use the same scoring rules in one plan.
Profile and patient readiness
| Parameter | Pillar | Points |
|---|---|---|
| Primary care category | Care relevance | 18 |
| Additional categories | Care relevance | 6 |
| Core services | Care relevance | 6 |
| Business name | Patient essentials | 5 |
| Branch address | Patient essentials | 5 |
| Entrance pin | Patient essentials | 3 |
| Weekly opening hours | Patient essentials | 7 |
| Special opening hours | Patient essentials | 3 |
| Accessibility information | Patient essentials | 7 |
| Branch website | Appointment access | 8 |
| Patient phone number | Appointment access | 7 |
| Appointment route | Appointment access | 7 |
| WhatsApp contact | Appointment access | 3 |
| Entrance and exterior | Visual usefulness | 5 |
| Interior and facilities | Visual usefulness | 4 |
| Clinic team | Visual usefulness | 3 |
| Treatment facilities | Visual usefulness | 3 |
Reputation
Patient ratings contribute 40 points, recent activity 20, routine review replies 25, and low-rating follow-up 15. Ratings use a 180-day half-life. Reply populations exclude reviews less than seven days old. A reply is not proof that a complaint has been resolved.
Listings health
We compare eligible directories against a confirmed clinic record. Formatting-equivalent names, numbers, and addresses count as matches.
AI visibility
The primary measure is the percentage of successfully collected eligible answers recommending the correct clinic. Each report identifies the tested queries, language, surface, model, and sample size. A controlled API answer is not presented as a direct observation of a consumer app.
Unread is never broken
Blocked pages, partial photo galleries, and incomplete review history remain unread. Confirmed missing information receives zero where applicable. Unsupported or ineligible fields are excluded with a stated reason.
Every score shows evidence coverage. Below 80% coverage, with unread essential identity, or with insufficient samples, we mark the assessment provisional. Mathematical complete-score bounds show the possible effect of unknown points; they are not statistical confidence intervals.
Initial weights, open to scrutiny
Version 0.1 uses proposed product weights and thresholds. Calibration against clinics and measured outcomes is still required. Each snapshot fixes its date, evidence, and rubric version so the same inputs reproduce the same score.
No Google account needed for public checks
Public checks use public information. Workspaces use their own sign-in. Connecting Google Business Profile is optional and required only to manage a clinic’s profile or reply on its behalf. Google Search Console access is not required.