Why your practice shouldn’t sound like ChatGPT

You can spot it now. “I hope this message finds you well.” “We truly appreciate your valuable feedback.” “We strive to provide exceptional care.” Nothing in those sentences is wrong, and that’s the problem — nothing in them is anyone, either. Your patients have read a hundred emails exactly like them this month, most written by the same handful of AI models, and they’ve learned to skim past that voice the way you skim past a parking-lot flyer.

Sameness is expensive

A practice isn’t chosen from a spreadsheet. Patients pick you — the doctor who explains things without hurrying, the front desk that remembers their kid’s name. Your voice, in every review reply and recall email, is how that relationship survives between visits. Generic AI is trained on everybody, so left to its defaults it writes like nobody. Every generic message you send spends a little of the thing that actually differentiates you, in exchange for saving a few minutes. Bad trade.

The fix is not avoiding AI. The fix is refusing its defaults.

How to teach a machine your voice

This is the method I use in paid builds, complete, no held-back step. It works in any decent AI tool.

1. Collect real writing. Five to ten things you or your staff actually wrote that sound like you — a patient email you were proud of, a review reply that got a warm response, even a text. Don’t polish them first. The flaws are the voice.
2. Write the rules. A short list of what you always do and never do. “We use first names.” “We never say ‘we strive.’” “We keep replies under four sentences.” “We thank people for hard feedback without getting defensive.” Ten lines is plenty.
3. Give the tool both, every time. Rules govern; examples calibrate. Paste them in with every request (or save them as standing instructions if your tool supports it), then ask for the draft. The difference in output is immediate and a little uncanny.
4. Keep the approval step. A voice-matched draft is still a draft. A human reads it, fixes what’s off, and approves the send. Over time, the fixes go back into the rules, and the drafts keep getting closer.

Rules govern. Examples calibrate. A human approves.

Why I’m giving away the recipe

Because knowing the recipe isn’t the hard part — the hard part is the patient, iterative work of building the voice file, wiring it into your actual tools, and training your front desk until it runs without anyone technical in the room. If you have the time, this guide is genuinely enough to do it yourself, and I’d be glad you did. If you don’t, now you know exactly what you’d be paying me for. Either way, your practice stops sounding like everyone else’s.