What does AI say about your clinic?
1 in 3 people in the Netherlands consulted AI about a health concern in the past year — often precisely because it is anonymous. Someone weighing up a hair transplant, an implant or laser eye surgery increasingly asks ChatGPT first. AgentReady measures — statistically and repeatably — whether your clinic appears in those answers, whether the facts are right, and whether an AI assistant can get anywhere on your website.
Source: Univé / PanelWizard, October 2025, n = 1,042
exampleclinic.nl
Your clinic: 0% of AI answers (CI 0–7%, n=48)
A fragment from an AgentReady report. The figures above are invented example values — never a real clinic’s data.
Four pillars, fixed weights.
One score (0–100), built from four pillars with fixed, published weights — score version 0.1.0.
Discoverability
30%Do AI assistants name your clinic on purchase-intent questions that mention no brand (“best clinic for…”)? Measured as a share of repeatedly sampled answers, set against the best-named competitor.
Discover
Factual accuracy
25%Are the claims AI makes about you correct? Every claim is checked against your own published pages. Incorrect pricing information carries extra weight.
Accuracy
Actionability
30%Can an AI assistant acting for a patient find prices on your site and submit an enquiry — or does it run aground on a CAPTCHA or an unreadable form?
Act
Accessibility
15%Is your site technically readable for AI systems: crawl access, structured data, and content that exists without JavaScript?
Access
A measurement, not a snapshot.
Sampled repeatedly, not asked once.
Every question is put to each engine several times (8× by default), across multiple engines: ChatGPT, Claude, Gemini and Perplexity. One answer is an anecdote; a sample is a measurement.
Every percentage carries a confidence interval.
Every percentage we quote comes with a 95% Wilson interval and the sample size. “Your clinic: 0% (CI 0–7%, n=48)” — that is how we report, always.
Dropouts are reported, never quietly dropped.
An engine that fails repeatedly is reported as reduced coverage — not silently left out. Every skipped comparison gets its reason in the report too.
Everything can be checked afterwards.
Raw answers are kept for every scan; scores can be recomputed offline. The score weights are versioned, so measurements stay comparable over time. Every measurement also carries its scan date: AI answers shift from day to day, so we never compare figures from different days as though they were one measurement.
The limits of the measurement, stated up front.
This field is full of promises no one can keep. That is why this is the most important part of our methodology.
No differences the sample cannot carry.
If the confidence intervals overlap, we do not call it a difference.
No promises of fast ranking movement.
AI visibility has two channels: what models “know” from training changes slowly; what searching engines such as Perplexity retrieve live can move faster. We report per channel and never promise a pace the channel cannot deliver.
No “caused by us”.
Improvements after an intervention are reported on a fixed attribution ladder — from “consistent with” to “plausibly attributable” — with competitor movement as the control. The phrase “proven cause” does not appear on it.
The method is public; the exact question wordings are not. Anyone who knows the exact questions can optimise for those sentences instead of for real visibility — and then the measurement measures nothing. The same model as standardised testing: method public, items private.
A measurement initiative by Nine Minds.
AgentReady is the measurement initiative of Nine Minds, the company of Ayça Gökalp. Nine Minds helps Dutch SMEs improve their processes before they automate — AgentReady applies that same level-headedness to the newest acquisition question: what happens when your next patient asks an AI assistant first?
Questions about the methodology are welcome: aycagokalp@nineminds.nl
Curious what AI says about your clinic?
We will send you the full report for your domain — including the measured percentages, the confidence intervals, and the factual errors AI is currently spreading about your clinic.
Request your reportWould you rather have 15 minutes of explanation first? Just email — no obligations.
Nine Minds · nineminds.nl · aycagokalp@nineminds.nl
Nine Minds (nineminds.nl) is a Dutch company, not affiliated with similarly named businesses abroad.