At your next appointment there is a reasonable chance your doctor will ask whether you mind an AI assistant listening in. It transcribes the conversation and drafts the clinical note, so that your doctor can look at you instead of a keyboard. Adoption moved quickly: around two thirds of US physicians report using or considering the technology, one large health system reports around 70 per cent of its physicians using it daily, and one health system alone recorded more than two and a half million encounters with it. Most patients say yes without knowing what they are agreeing to. Here is what is worth knowing.
The software listens to the consultation, transcribes it, and drafts a structured clinical note in the format doctors use: your history, the examination, the assessment, the plan. Your doctor reviews and edits that draft, then signs it into your record.
Three points people commonly get wrong:
Electronic records created an enormous documentation burden, and a great deal of it spilled into evenings and weekends. Time studies of these tools report meaningful reductions in note-writing time, in the order of quarter-hours per day, which sounds small until you multiply it across a career.
The effect patients notice is different and more human. The thing clinicians mention most often is eye contact. A doctor who is not typing is a doctor who is looking at you, and who is more likely to catch the hesitation before you mention the symptom you were not going to bring up.
You should be asked. If you are not, you can ask directly: is an AI scribe being used, is the audio retained, and who else can access it.
You can decline, and your care should not change because you did. Reasonable grounds include a consultation about something highly sensitive, an unresolved legal or employment matter, or simple discomfort. It is your appointment.
Also worth asking, if the answer matters to you: is the vendor processing the audio outside your country, and is the transcript covered by the same protections as the rest of your medical record.
This is the practical advice most patients never receive. If you have access to your visit notes through a portal, read them.
These systems are accurate enough to be useful and imperfect enough to need checking. Reported problems include misattributed statements, hallucinated detail that was never said, mistakes with medication names and doses that sound similar, and difficulties with strong accents or consultations conducted partly in another language. A speech-based system that mishears "hypertension" for "hypotension" produces a note that reads plausibly and says the opposite of the truth.
What to check:
If something is wrong, say so. You can request a correction, and it is far easier to fix now than to find out two years later that an error has been quietly copied forward into every subsequent note.
The research so far is more encouraging on time than on accuracy, and both deserve a fair hearing.
On time, the findings are consistent: documentation time falls, and clinicians report less after-hours work and lower burnout. On experience, physicians report being able to attend to the patient rather than the screen, and patients report the impact on their visit as positive or neutral rather than negative.
On accuracy, the picture is more mixed and still being studied. Notes drafted this way are generally complete and well structured, and the errors that occur tend to be errors of the confident kind: a plausible sentence that was never said, a medication name that sounds like the right one, a symptom recorded as present when it was denied. Those are harder to spot than obvious nonsense, which is precisely why review matters and why your own read of the note is worth the five minutes.
There is also a documented gap in performance for accented speech and multilingual consultations. If your appointment moves between languages, or if you have been misheard by voice software before, it is reasonable to check the note with extra care.
Notes written this way tend to be longer and more complete than typed ones, which is mostly good and occasionally overwhelming. If you are managing a long-term condition across several clinicians, the volume of documentation you now receive is more than most people can hold in their heads.
That is an argument for keeping your own copy rather than relying on a portal you log into twice a year. If your after-visit summary is dense with terminology, our medical report analysis will put it into plain language, and a Health Passport keeps visits, results and medicines together so the next appointment does not start from memory.
An AI scribe documents the consultation you had. It does not make a rushed appointment longer, it does not make a dismissive clinician attentive, and it does not improve a diagnosis.
If you have ever left an appointment feeling you did not get your concerns across, the tool that helps is preparation: knowing the order of your symptoms, what makes them better or worse, and what you want to leave with. Working through a structured symptom check beforehand is a way of doing that thinking while you are calm rather than in the room.
Ambient AI scribes are now common in consultations, they free your doctor from the keyboard, and they draft a note your doctor must still review and sign. You can ask whether one is in use, what happens to the recording, and you can decline. The single most useful thing you can do is read the note afterwards and check your medicines, allergies and plan, because an error that enters your record tends to stay there.
This article is for general education and is not a substitute for professional medical advice. Policies on AI documentation, consent and data retention vary by provider and country. Ask your own clinic what applies.
From diagnosis and drug discovery to patient triage and personalised treatment, AI is reshaping every corner of healthcare. Here is where things stand in 2026 and what comes next.
Taking multiple medications increases the risk of dangerous drug interactions. Learn what drug interactions are, which combinations to watch for, and how AI tools can help you stay safe.
We've redesigned our AI thinking indicator to show you exactly what's happening as our engine analyzes your symptoms — no more generic loading spinners.