AI Disclosure

Where AI is used

• Classification. Assigning each public item to themes from a reviewed clinic taxonomy, and detecting sentiment and risk signals.
• Summarisation. Writing plain-language summaries of what a group of public items says.
• Response drafting. Producing a draft reply for a human at the clinic to review, edit, and post themselves.
• AI visibility measurement. Sending your defined prompts to answer engines and recording what comes back as evidence.

Where AI is not used

• The Trust Health Score. It is computed by a deterministic, versioned formula. No model produces, adjusts, weights, or overrides the number.
• Coverage Confidence. Determined by explicit thresholds over evidence volume and source diversity.
• Risk-state transitions. Driven by defined rules over the deterministic measurements.
• Access, billing, and entitlement decisions. None of these involve a model.
• Publishing. Nothing is posted anywhere on your behalf.

Why the split matters

A reputation number that changes because a model behaved differently on a second run is not a measurement. Keeping the model on the classification side of the line means the same inputs and the same formula version always produce the same score, the methodology can be published in full, and a movement in the number is a movement in the world.

Human oversight

Classifications are visible and challengeable: every theme opens onto the public items behind it, so a misclassification can be seen rather than inferred. Response drafts always require a person to review and post. No automated decision produces a legal or similarly significant effect on any individual.

What is sent to AI providers

Only public source material and the prompts you define. Patient data is never sent because it is never collected. Workspace credentials, billing details, and support correspondence are never sent. The AI providers currently used are listed in Subprocessors.

Limits you should assume

Classification is good, not perfect; summaries can flatten nuance; and an answer engine’s description of your clinic is a record of what it said on a date, not a statement of fact. Every AI-derived output in the product is presented with the evidence behind it precisely so you can check it.