MedicalScribe.app

Note writing is the whole business. Medical Scribe records a visit or a dictation, transcribes it with speech models it says are trained on medical terminology, and summarizes it into a structured note in the clinician's chosen template, from a SOAP note to a physical therapy progress report or a traditional Chinese medicine treatment record.

The apps for web, phone, Mac and Apple Watch and the Chrome side panel exist to capture audio and deliver the note. With the models removed there would be nothing to sell, and the company competes directly with other AI scribes, setting itself against Commure, Freed, Heidi, PatientNotes and Twofold.

Every note is reviewed and edited by the clinician before it is copied into the record, on a phone, a computer or the Chrome side panel beside the EHR. Nothing is written to a chart by the product itself, so the clinician controls what enters the record.

The review step gets no help from the product: nothing links a line of the note to the audio it came from, and nothing flags passages transcribed with low confidence, which matters for a tool promising the same accuracy in 57 languages. Medical Scribe offers a client consent form so clinicians can record with the patient's agreement.

The approach is described in general terms: speech to text models the company says are designed and trained for medical terminology, followed by AI summarization into the chosen template. Examples of what the models handle, drug names with doses, diagnoses and anatomical terms, are shown on the home page.

No model family, provider, version, training data or update practice is identified, and it is not clear whether the speech models are built in house or adapted from another provider. Clinicians cannot tell when the models behind their notes change.

The architecture is described as the company's own speech to text models followed by AI summarization, and no model provider, speech service or cloud host is named. Whether the summarization runs on an outside language model is not stated, and Leveled Platforms, Inc., the company behind the product, is the only party named anywhere in the chain.

No subprocessor list exists, and the business associate agreements the company signs are not posted. A clinician therefore cannot see which companies receive patient recordings, or on what retention and training terms, before using the product.

One named clinician and a user count are the evidence. Dr. Sadaf Islam of Innovatel Telepsychiatry calls the app a game changer, and the company says it is loved by more than 5,000 clinicians, with no figure for time saved, note accuracy or how often notes are edited.

The accuracy language is strong, industry leading accuracy for medical terms and unmatched precision for drug names and procedures, and no word error rate, test set or comparison backs it. A clinician has the free trial to judge transcription quality on their own visits.

Assurances are general: end to end encryption, HIPAA compliance and unmatched privacy and security standards. Nothing states how long recordings, transcripts or notes are kept, whether clinicians can delete them, whether they are used to train the speech models the company describes as its own, or how data is de identified.

Recordings can start on an Apple Watch and sync to phones and computers, so patient audio moves across several devices and accounts, and nothing explains what is stored where. Safety engineering is not addressed beyond the clinician's review of each note.

HIPAA compliance is claimed throughout, and business associate agreements are listed among Medical Scribe's compliance measures. Nothing describes the agreement's terms, which plans include it, or whether it binds the speech and model providers behind the product.

The privacy policy covers only website visits, logging, advertising cookies and text message consent, and says nothing about recordings, transcripts or notes. Encryption and HIPAA safeguards are described in marketing terms. A clinic has a stated willingness to sign an agreement and no written document behind it that reaches patient data in the service.

No attestation is held yet: SOC 2 Type II is listed as in progress, without a target date or auditor. The company says it runs regular security assessments and penetration testing, without naming who performs them or how often.

End to end encryption and HIPAA safeguards are described in marketing terms, and a clinic has no trust center or document to request. Until the SOC 2 examination is complete, a clinic has only the company's description of its controls.

Documentation is the whole pitch, and no regulatory position is stated. Medical Scribe describes itself as an AI assistant that saves clinicians hours by transcribing patient interactions and dictations into notes ready for the record, which is the nearest thing to a statement of purpose.

It transcribes and summarizes what the clinician and patient said, producing notes the clinician reviews and copies into the record. Nothing markets it as diagnosing, recommending treatment or advising patients, and the templates format the clinician's own visit. The documentation positioning matches the product, with no intended purpose statement to set that limit in writing.

Nothing describes how the speech models or note generation are tested or governed, while recognition quality across languages is a headline claim: Medical Scribe says its models support 57 languages with the same medical terminology accuracy as English. No results by language, accent or specialty back that parity claim.

A clinician seeing patients in Spanish, Portuguese or any of the other supported languages cannot tell how often the transcript or note goes wrong, and a mistranscribed drug name or dose in a second language is the kind of error a busy reviewer misses.

Review before copying is the mechanism: the clinician reads and edits every note before it reaches the record. Nothing else stands behind the output.

Transcription and summarization carry no error rate or warranty, and the accuracy claims, industry leading for medical terms and the same across 57 languages, come with no remedy if they fall short. The terms of service set the limits on liability. A clinician who finds the notes unreliable can stop at the end of the trial or billing period, and every signed note remains the clinician's own responsibility.

No record system integration exists. The Chrome extension keeps Medical Scribe open beside a browser based EHR, and the clinician uses Copy to move the finished note into the chart by hand; desktop EHRs and phone users copy and paste the same way.

What Medical Scribe calls EHR integration is that side panel rather than any connection to a named EHR, and nothing is read from the chart or written to it automatically. For individual clinicians this keeps setup instant; for clinics it means no data flows between the scribe and the record.

Nothing says where Medical Scribe runs or where recordings and notes are stored. It is a hosted service reached through web, iPhone, Android, Mac and Apple Watch apps, with no cloud provider, region, tenancy model or on premises option named. A recording can begin on a watch and finish on a phone or computer, so patient audio passes between several devices before it reaches the service.

Support for 57 languages suggests users outside the United States, and no statement addresses whether their data, or US clinicians' data, stays in a particular country, or where transcription happens.

Pricing is listed and the product can be started alone. The individual plan is $42 per user per month billed yearly at $499, with monthly billing available at a higher rate, unlimited patient visits and a standard free trial; a 30 percent discount code for new clinicians runs on the site.

Teams, hospitals and clinics get volume pricing on request. Nothing is listed for setup or implementation, and the free trial and individual plan can be bought from the website without speaking to anyone.

Coverage is broad and named by specialty: acupuncture and traditional Chinese medicine, cardiology, direct primary care, family medicine, physical therapy and psychology, plus veterinary records, with templates for each and a claim of use in hospitals and clinics on enterprise terms.

The one named user practices telepsychiatry, and no clinic, hospital or specialty group is named as a customer. The breadth rests on templates and specialty pitches, and the support for 57 languages widens the claim further without a deployment that shows it working in a given setting or language.