If you go looking for a doctor visit summary app for families, you mostly find recorder apps. Medcorder, Kin, Doctor Notes, DocNotes AI, AI Doctor Notes β most of the category leans on the same primitive: record the appointment, transcribe it, hand the family a searchable text of what was said.
It's a real improvement over the previous default, which was nothing. Before recorder apps, a family member who wasn't in the room got whatever their sister or brother could remember to text in the parking lot β usually a fraction of what mattered, sometimes wrong, often missing the one detail the absent sibling cared about most. The recording closes that gap mechanically: now there's a primary source. Anyone can play it back.
The trouble is that anyone can play it back turns out to be roughly equivalent to no one will play it back. The audio file is 47 minutes long. It has dead air. It has the doctor leaving the room for five minutes to look something up. It has the patient asking about the wrong leg. The transcript is more skim-able, but a 9,000-word transcript of a 47-minute appointment is still a 9,000-word document the brother in Cleveland will open, look at for ten seconds, and close.
A recording is the source, not the output. The output a family caregiver actually needs β the artifact they can hand to the sister in Seattle, the cardiologist's referring office, the lawyer drafting the POA β is a summary. And summarization is a different and harder skill than transcription.
What a useful visit summary contains
The same four pieces, every time, in a format short enough to read on a phone in line at the pharmacy:
What changed. The new diagnosis, if there was one. The new medication, with the dose and the schedule. The discontinued medication, with the reason. The lab value that moved enough to matter. Changed is the load-bearing word β the family doesn't need a transcript of everything the doctor said, they need the diff from the last visit.
What to do next. The appointment to schedule, the test to get done before next time, the medication to start tomorrow, the symptom to call about. Action items with owners. Not "Dr. Patel discussed the importance of titrating slowly," but "Mom takes 5mg of the new pill tonight; if no side effects, increase to 10mg in a week."
What's still open. The questions the family didn't get to ask, the test result still pending, the specialist referral still in progress. The visit summary that doesn't name what's still open creates a false sense of completion β and the family member reading at 9pm has no way to know what they should be asking about between now and the next visit.
What to ask next time. The follow-up questions that surfaced during the visit but couldn't be answered in the time available. "Ask about the prednisone taper schedule." "Ask whether the kidney function trend means we should be worried." "Ask if there's a Spanish-language patient handout for this."
That's it. Four sections, usually under 300 words total. It's the kind of summary a sibling in Cleveland will actually read on the way to lunch β and the kind a new specialist can scan in the 60 seconds before they walk into the exam room.
Producing that summary from a 47-minute recording is the work the recorder apps don't quite do. They get close β most of them now offer an AI-generated summary alongside the transcript β but the summaries we've seen tend to be either bullet-pointed transcripts (faithful to what was said, useless for what to do next) or hallucinated optimism (smoother than the source, but inventing structure that wasn't there).
Why this is harder than it looks
Summarization is a judgment task. The right summary depends on who's reading it β a sibling needs the action items, a referring specialist needs the medication changes, an attorney needs the capacity assessment, a hospice nurse needs the comfort-care preferences. A general-purpose AI summary tries to be all of these and ends up being none of them.
The other half of the difficulty is that good summaries acknowledge the things the appointment didn't resolve. The model that's been trained on "produce a confident summary" will paper over the open questions. The model that's been prompted to mark its own uncertainty β the doctor mentioned a possible referral but didn't commit to one β produces summaries that are harder to misread.
We don't think the recorder apps are wrong to record. The recording is a useful primary source β it's the only place to go back to when the summary turns out to have missed something, and the act of recording often makes the patient and the doctor more careful about what they say. What we think is that the recording isn't the deliverable, and treating it as one shifts the summarization work back to the family member who couldn't be at the appointment in the first place.
What we built
Kintaria's visit summary feature takes a transcript, recording, or pasted notes and produces the four-section summary above β what changed, what to do, what's still open, what to ask next time β with the medication changes structured separately so they flow into the shared medication list without retyping. The original transcript is preserved as the primary source; the summary is the artifact the family actually reads.
When the family caregiver wants to send the summary outside the workspace β to a new specialist, a hospital social worker, an estate attorney β the share-with-provider link generates a time-limited, scope-limited view that includes only the sections the caregiver chose, expires automatically, and can be revoked instantly. The recipient doesn't need a login. They see the summary in 30 seconds, not 47 minutes.
If the patient reads more comfortably in a language other than English, the bilingual workspace shows the same summary side-by-side in the parent's language β so the mother who couldn't follow the cardiologist's English in real time can read the four-section summary in Mandarin or Korean or Vietnamese at the kitchen table the next morning. The summary is for the family, but it's also for her.
The honest comparison
If your hardest problem is having a verbatim record of what the doctor said β for a malpractice question, a second opinion, or your own peace of mind β a dedicated recorder app like Medcorder does that job well, and we don't try to compete with it on the recording itself. We think of recorders as an upstream tool, the way you might think of a high-quality microphone separately from the podcast you produce with it.
If your hardest problem is that the sibling in Seattle keeps asking what happened at the last appointment, that the new specialist needs the medication history before the visit, that the parent can't follow the doctor's English in the room, or that the family keeps having the same conversation about what the cardiologist said in March β those are summary problems, not recording problems. They are what we built for.
The recording is the source. The summary is the artifact. They are not the same thing, and the family member you're trying to help β the brother in Cleveland, the sister at 9pm, the lawyer at next Tuesday's signing β needs the second one.