Key points
- In a review of 33 studies, 76% found at least one outcome better when patient and physician shared a language other than English.2
- 67.8 million people in the US spoke a language other than English at home in 2019, and 62% of them spoke Spanish.1
- Professional interpreters made errors of potential clinical consequence 12% of the time, against 22% for ad hoc interpreters, in Spanish-language paediatric emergency visits.3
- Speech recognition accuracy varies by speaker, so the clinician reviews every line.4 wai understands more than thirty-five languages, and the clinician checks and signs every draft.6
Summary
An AI scribe drafts the note from a consultation held in Spanish or another language it understands, and the clinician reviews every line before signing. With an interpreter in the room the scribe hears three voices, so attribution is the first thing to check.
The conversation stays in the language that serves the patient best, and the clinician still leaves with the note drafted. Language-concordant care improves outcomes, and a scribe that understands the consultation as it happens removes the work of translating the record afterwards.
See it in your clinicThirty minutes on your own patients and doctors.
Request a demoWatch a consultation held in Spanish drafted as a note, with the patient's own words kept where they matter, ready for your signature, in fifteen minutes.
| Consultation set-up | What the scribe hears | What the clinician checks before signing |
|---|---|---|
| Clinician and patient share Spanish or another language | Two voices, one language | Medicines, doses, allergies and negatives, read against the visit |
| Professional interpreter, in person or on the call | Three voices, two languages | Interpreter named, and that answers are attributed to the patient, not the interpreter |
| Family member interpreting | Three voices, with paraphrase likely | Any clinical detail heard only through the relative, flagged and confirmed |
| Patient switches language mid-sentence | Mixed speech, the hardest case for recognition | Symptom descriptions and timings, line by line |
| Patient's own words matter clinically | The original phrase | The phrase kept in quotation marks where it carries meaning |
| Noisy setting or poor line | Gaps and guesses | Every gap marked for the clinician to complete |
The patients are already here
The US Census Bureau reports that 67.8 million people, almost one in five, spoke a language other than English at home in 2019, and that Spanish was the most common, at 62% of them.1 Gulf clinics work in Arabic and English side by side, and European clinics see patients from several countries, so the note has to follow the language of the room.
The shared language wins
A 2019 systematic review in the Journal of General Internal Medicine examined 33 studies of patients with limited English proficiency. Seventy-six percent found at least one outcome better for patients receiving language-concordant care, and 15% found worse outcomes.2 None of the studies formally assessed the clinicians' language skills.2
A genuinely fluent clinician consults in the patient's language, and the clinic keeps a clear record of who is fluent. Where nobody is, a trained interpreter joins.
Interpreters, and what the scribe hears
A study of 57 audiotaped visits with Spanish-speaking families in the two largest paediatric emergency departments in Massachusetts counted 1,884 interpreter errors, of which 18% had potential clinical consequences.3 The share of consequential errors was 12% with professional interpreters, 22% with ad hoc interpreters and 20% with no interpreter.3 Professional interpreters with at least 100 hours of training made a median of 12 errors, against 33 for those with less training.3
The scribe records the words spoken, often in two languages, so an interpreting error reaches the draft unless the clinician catches it. Use trained interpreters and name them in the note.
Transcription accuracy, measured
A 2020 study in the Proceedings of the National Academy of Sciences tested five commercial speech recognition systems on 19.8 hours of interviews and found an average word error rate of 0.35 for Black speakers against 0.19 for white speakers, a gap the authors traced to the acoustic models.4
That study was in English and covered the category as a whole. In any language, accents and switching languages mid-sentence move accuracy, and a drug name or dose is where an error costs most.
Over 12 months across a hospital network in Spain, an ambient scribe was used in 1,032,558 emergency consultations, with mean transcription accuracy of 93.9% and higher audited documentation quality.5 At 93.9% there are still words to check, so every note is reviewed before it is signed. The guide to scribe accuracy sets out how to test it.
How wai handles other languages
wai understands more than thirty-five languages and is built for Arabic-first care in the Gulf.6 A consultation held in Spanish, Arabic or another supported language is drafted as a note during the visit, and the clinician reviews, edits and signs it before it enters the record.
Templates are shaped to the clinician's specialty and refined as they edit, so the patient's own words for key symptoms can be kept in the note as a rule.7
Where to start
Pick ten recent consultations held in the language you most often use besides English. Run them through the scribe in a trial and read every draft against the visit, marking errors in medicines, doses, negatives and timings apart from wording.
Then set two rules. Interpreted consultations name the interpreter and the language in the note, and any medicine, dose or allergy heard in another language is read back to the patient before signing. Tell patients about the scribe in their own language, using the wording in the patient consent guide.
Bring ten consultations in your second language, and read the drafts with us.
Frequently asked questions
Can an AI scribe work in Spanish?
Yes, if it understands Spanish. wai understands more than thirty-five languages.6 The clinician still reviews and signs every note.
How accurate is AI transcription in other languages?
It varies by speaker, accent, noise and language mixing. Across one hospital network in Spain, mean transcription accuracy was 93.9% over 12 months.5 Test it on your own consultations before relying on it.
Can the scribe run with an interpreter in the room?
Yes. Name the interpreter and the language in the note, and check that the patient's answers are attributed to the patient. Professional interpreters made fewer consequential errors than ad hoc ones in one Spanish-language study, 12% against 22%.3
Should a bilingual clinician consult in the patient's language?
Where they are genuinely fluent, the evidence favours it. Seventy-six percent of 33 studies found at least one better outcome with language-concordant care.2
Does the scribe replace an interpreter?
The scribe drafts the note for the clinician, and a trained interpreter carries the conversation when no one shares a language.
How are patients told about the scribe in another language?
In their own language, in plain words, before the history starts. Say what the scribe does, that the clinician checks every word, and that they can ask for it to be switched off.
Sources
- 1US Census Bureau, 2022Nearly 68 Million People Spoke a Language Other Than English at Home in 2019Read 14 October 2026Read source
- 2Journal of General Internal Medicine, Diamond and colleagues, 2019A Systematic Review of the Impact of Patient-Physician Non-English Language Concordance on Quality of Care and OutcomesRead 14 October 2026Read source
- 3Annals of Emergency Medicine, Flores and colleagues, 2012Errors of medical interpretation and their potential clinical consequences: a comparison of professional versus ad hoc versus no interpretersRead 14 October 2026Read source
- 4Proceedings of the National Academy of Sciences, Koenecke and colleagues, 2020Racial disparities in automated speech recognitionRead 14 October 2026Read source
- 5International Journal of Medical Informatics, Alcazar-Peral and colleagues, 2026Deployment of an ambient AI scribe in emergency care: A 12-month evaluation in a large Spanish hospital networkRead 14 October 2026Read source
- 6wai, 2026AI scribes in the UAE, what a clinic must askRead 14 October 2026Read source
- 7wai, 2026wai for cliniciansRead 14 October 2026Read source
Consult in the patient's language, and leave with the note ready to sign.
Fifteen minutes on a consultation in your second language with a clinician who uses wai every day, and a reply to your request within one working day.
Partner with us
See it on your own patients, then build it with us
Thirty minutes on your own patients, doctors and rules, in your market. We reply within one working day, and the clinics that partner with us early shape what we build next.
Read how multilingual care works in a clinic