AI visibility measurement

Patients are asking AI which dentist to trust.

FIDESTIS measures what ChatGPT, Gemini, Claude and Perplexity actually say when someone asks who to see for implants or veneers in your city — and shows you which sources those answers are built from.

40Questions

4AI assistants

5Intent stages

384AI answers analysed

Fixed fee·7-day delivery·Raw data available·Spot-checked by hand

Figures from the August 2026 Sydney panel. Your report shows your own city, your own numbers.

How the measurement works

A fixed panel of patient questions Forty questions phrased the way patients actually phrase them, across five intent stages — from “how much do implants cost” to “who should I go to.”

Asked across four assistants Every question is put to ChatGPT, Gemini, Claude and Perplexity with live web access, then timestamped and stored. A sample is re-asked by hand in the consumer apps, because those surfaces don’t always match the APIs.

Every recommendation recorded Which clinics each assistant names, in what order, how strongly, and which websites it cited to get there.

Scored against your market You get a visibility score, a ranked leaderboard of the clinics in your city, and the citation-source table that explains why they appear and you don’t.

Fig. 1 — Panel procedure

The four assistants don’t read the same sources

One directory was cited 145 times by ChatGPT and never once by Claude, Gemini or Perplexity. Facebook was cited 65 times by Gemini and never by ChatGPT. Checking one assistant tells you about one assistant.

Source	Total	ChatGPT	Claude	Gemini	Perplexity
whatclinic.com	155	13	44	12	86
bestinhood.com	145	145	0	0	0
facebook.com	115	0	45	65	5

Fig. 2 — Sources cited, Sydney panel, 26–27 August 2026. Three of 796 distinct sources recorded across 384 answers. Individual practice websites are excluded from this sample.

Who runs it

I’m Eugene Naude. FIDESTIS is me — one person who builds and runs the measurement, writes the report, and takes the call.

I built this because the shift to AI-answered search is measurable and almost nobody is measuring it at practice level. Analytics can’t see it: there’s no click to record when an assistant recommends a clinic. So I built the instrument that can.

I work with a small number of practices at a time, in Australia and the UK.

eugene@fidestis.com

The audit

$1,900 AUD

Fixed fee · Delivered in 7 days

Your visibility score across all four assistants, with the raw data behind it
A ranked leaderboard of the clinics competing with you in AI answers
The citation-source table — the specific sites each assistant drew on
What the assistants say when a patient asks about your practice by name
A prioritised 90-day roadmap your web team can execute from

Request an audit

Or ask for the one-page snapshot for your city first — no charge, no obligation.

Most practices start here. If you want the panel re-run monthly and help executing the roadmap, that’s available afterwards — but the audit stands on its own.

Why not just check ChatGPT yourself?

You can, and you should — it takes a minute. What one check can’t tell you is whether that answer is typical, how the other three assistants differ, or which sources produced it.

In this panel, one directory was cited 145 times by ChatGPT and never once by the other three. A single check on a single assistant on a single day is a data point. Forty questions across four assistants is a measurement.

What I don’t claim

No one can guarantee that an AI assistant will recommend a particular practice, and anyone who tells you otherwise is selling something they can’t deliver. What I sell is measurement, a diagnosis of where the answers come from, and a roadmap.

The same panel is re-run on request, so movement is measured rather than asserted — and AI answers do vary between sessions, which is why every figure in your report is tied to a timestamped response you can inspect.