Clinicians keep asking the same questions about AI medical scribes, just not in the keyword form a search engine expects. They ask the way you’d ask a colleague: does this store my patients’ audio, do I need consent, will it mess with my billing, what does it cost. This page answers the ones we hear most, in plain language, each one short enough to quote and each one linked to a deeper post if you want the long version.
One ground rule before we start. We answer for AI Medical Scribe by Patient Square specifically, and where a question is really about the whole category, we say “ask every vendor.” A few answers concede where we don’t fit. That’s deliberate. A FAQ that pretends the product is perfect isn’t a FAQ, it’s an ad.
Key takeaways
- Visit audio is processed in memory and discarded once the note is drafted. No archive, for anyone.
- We suggest ICD-10 codes only. We do not set your E/M level, and that protects you from upcoding exposure.
- Launch pricing is $89/clinician/month, on annual billing, with a 7-day trial.
- We capture English and more languages; the note always comes back in clean clinical English.
- Our SOC 2 Type II audit is underway, and we don’t call it “certified” until it’s done.
to review the draft note after the visit ends
US launch price per clinician (annual billing)
of visit audio kept after the note is drafted
AI Medical Scribe by Patient Square is an ambient AI medical scribe that listens during the visit and hands back a structured SOAP note, ICD-10 suggestions, and a prescription draft, ready to review and sign about two minutes after the visit. Everything below is a question about some part of that sentence.
Privacy, consent, and your data
These come up first in almost every demo, and they should.
Does an AI scribe store my patients’ audio?
It depends on the vendor, and you should ask each one directly. Some retain recordings for days or weeks, some let you opt out, some won’t give you a straight number. Ours doesn’t keep it: visit audio is processed in memory and discarded the moment the note is drafted, so no archive exists. The full breakdown, with competitor policies quoted and dated, is in our audio-retention comparison.
Is an AI medical scribe HIPAA compliant?
Software by itself is never “HIPAA certified.” That certificate doesn’t exist, so be wary of anyone who implies they hold one. What actually matters: does the vendor sign a BAA, encrypt PHI in transit and at rest, scope access by role, and tell you where the audio goes. We map our safeguards to the HIPAA Security Rule and sign a BAA with every US customer.
Do I need patient consent to record, and is it state-specific?
Yes, and yes. US recording-consent law splits by state: some need one party to agree, others need everyone in the room. Get a clear verbal yes before capture starts and note it in the chart. Cross a state line and the rule can change. We keep a state-by-state consent breakdown so you can check yours before you start.
What do I tell the patient before I start recording?
Keep it short and routine. Something like: “I use a secure tool that helps me write my notes so I can focus on you instead of the screen. It’s fine with you?” Most patients say yes and appreciate the eye contact. For telehealth, say the same thing and get the yes on the call. If a patient declines, you turn it off and chart the usual way.
Is my patient data used to train your AI?
No. We don’t train our model on your visits, and we don’t sell or share your clinical data with anyone. The structure of the product makes this easy to promise: since audio is discarded once the note is drafted, there’s no recording to feed into a training set later. What remains is the note you signed, and that note is your practice’s property. You can export or delete any visit whenever you want.
How is the data encrypted and stored?
Notes are encrypted in transit with TLS 1.2 or higher and at rest with AES-256. Access is role-scoped and logged, so you can see who opened what. Notes belong to your practice, not to us. The whole posture sits on our security page and in our security checklist.
How it works and what it hands back
How fast is the note?
About two minutes after the visit ends. The draft is ready while the patient is still leaving, not waiting for you that night. You read it, correct anything off, and sign. That speed is the whole pitch of an ambient scribe: the chart closes when the visit closes. More on the mechanics in what an AI medical scribe is.
Does it write prescriptions?
It writes a prescription draft. It does not e-prescribe, send to a pharmacy, or touch SureScripts. The draft saves you typing; you review and sign like everything else. Critically, every Rx draft runs through a deterministic safety screen before you can sign, which we explain next.
What’s the Rx safety screener?
It’s a rule-based check, not a guess from the language model. Each prescription draft passes through drug-interaction, renal-dosing, and pregnancy checks, then re-screens at sign time so a change you make can’t slip past it. Unsafe combinations are hard-blocked; you can override, but only with a recorded attestation. We built it that way because a draft an LLM wrote is still a draft an LLM wrote, and the safety layer shouldn’t be probabilistic.
Will an AI scribe change my E/M level?
Not ours, by design. We surface ICD-10 diagnosis suggestions you review and confirm. We do not set your E/M visit level, generate CPT or HCC codes, or file claims. Tools that auto-assign an E/M level take on the upcoding and audit risk that’s legally yours. We think that’s a line worth keeping, and we explain why in ICD-10 suggestions versus a coding engine.
How accurate is it?
We don’t publish an accuracy percentage, and we’d distrust any vendor who waves one around without saying how it was measured. Drafts contain errors: a misheard drug name, two complaints compressed into one. That’s exactly why review-and-sign is load-bearing and not optional. We wrote an honest piece on what accuracy actually means for a scribe and how to test it on your own visits.
Languages, connectivity, and your setting
What languages does it support?
English and more languages, with the note always returned in clean clinical English. The honest advice is to read any vendor’s language list literally, then test it on your own patients during a trial. Accents and code-switching are where models quietly break, and “multilingual” without a named list is marketing, not a feature.
Does it work offline or on low connectivity?
Yes. Capture works offline with on-device AES-256-GCM encryption, then syncs when the connection returns. That matters for any setting where the link drops mid-visit. The patient conversation isn’t lost, and the encryption travels with the captured data. Full detail in offline and low-connectivity capture.
Does it work with my EHR?
It works with any of them, because it doesn’t integrate with any of them. You export the structured note and bring it into Epic, athenahealth, eClinicalWorks, Elation, or whatever your clinic runs. No live write-back, no integration fee, no waiting on an IT ticket. That EHR-agnostic design is a feature, and we contrast it with free EHR-bundled scribes.
Compliance and certifications
Is it SOC 2 certified?
Not yet. Our SOC 2 Type II audit is underway, and we won’t call ourselves “certified” until it’s done. Be careful with vendors who blur “audit in progress” into “certified,” because those are different things. What we can offer today is the HIPAA Security Rule mapping and a BAA, covered in our HIPAA and BAA guide.
What happens if I want to leave?
Your notes are your practice’s property, full stop. You can export or delete any visit yourself, with no support ticket and no per-record fee, and we never sell or share clinical data. Export friction is how vendors keep you, so read the contract on this point before you sign, not after. Detail in data ownership and export.
Fit: specialty, segment, and the honest limits
Is it good for behavioral health or therapy?
The zero-audio-retention design fits behavioral health well, because no recording of a therapy session ever persists. That’s the single biggest privacy worry in this space, alongside 42 CFR Part 2 and psychotherapy-note sensitivity. Read the psychiatry and privacy post for the specifics before you decide it fits your practice.
Does it work for my specialty?
It works across specialties, but we won’t pretend a generic scribe is tuned to every one of them. We don’t ship per-specialty note templates today. The honest test is to run it on your own case mix during the trial and read the drafts closely. We have focused write-ups for primary care, pediatrics, and nurse practitioners.
How does it help with insurance claims and denials?
A complete, structured note at the time of the visit is the documentation a payer wants to see. Broadly, that means fewer claim denials traced to thin documentation. The scribe doesn’t file or code the claim. It makes the underlying record defensible, which is upstream of every denial fight.
How much does it cost, really?
Plain numbers, since this is where most evaluations end up.
| Plan | Per clinician / mo (annual billing) |
|---|---|
| Solo | $89 |
| Group (most popular) | $79 |
| Trial | 7-day free |
These are annual-billing launch rates; month-to-month and the full ladder sit on the pricing page. Every tier is the same product, with no feature gating between Solo and Group.
For comparison, published US self-serve scribes run roughly $39 to $199 a month, and several large vendors (Suki, DeepScribe, Nabla) publish no price and route you to sales. Transparent list pricing makes a vendor easier to evaluate without a sales call. We lay out the full rate card in our pricing breakdown.
Does it pay for itself? Primary-care physicians in a 2023 JAMA Network Open study logged a median of 36.2 minutes of EHR time per 30-minute visit. Against an $89 monthly line item, recovering even part of that isn’t a close call. Run the math on your own visit volume, though, not ours.
Which scribe should I actually pick?
We’re one option, and the honest answer depends on your constraints. If you need deep bidirectional EHR integration or an enterprise health-system rollout today, a different vendor likely fits better, and we’ll tell you that on a call. If your priorities are zero audio retention, flat transparent pricing, ICD-10 and Rx drafts you control, and a BAA, we’re built for exactly that.
The way to know is to run it on a real clinic day. Take the 7-day trial, use it on your own patients, your own accents, your own interruptions, and read every draft for the first week. That’s the evaluation that settles it. When you’re comparing options, our how-to-evaluate guide and the best-scribes roundup lay out the questions to ask everyone. Then book a demo and bring your hardest case.