4,500 patients linked their charts to ChatGPT. One US health system gave itself two months to respond
Published · By Gaurav Phogat
Ochsner Health’s CIO says about 4,500 of its patients have linked their records to ChatGPT, and that the system must decide its response within roughly two months. India has no MyChart, but I think our hospitals face the same decision in a messier form.
What happened
healthsystemCIO published on 28 October 2026 remarks made on 30 September by Amy Trainor, CIO at Louisiana’s Ochsner Health. She said about 4,500 Ochsner patients had connected their records to ChatGPT by that date.
New to the topic? Industry Watch: the themes so far sets out the reading order.
Her options, as reported: expand Emmie, Epic’s patient-facing assistant, which Ochsner runs for a small group; invest in its own internal model process; or do neither, which she warned would mean losing patients to ChatGPT. She also raised opening appointment availability to consumer AI tools, though she said the AI companies were not yet ready for that conversation, and she questioned whether paid placement in AI answers could compromise impartiality.
Separately, Becker’s reported on 21 October that patients linking MyChart accounts to ChatGPT at Advocate Health rose more than 7,000% from July to August.
My take
Her third option makes the point: doing nothing is still a decision. Patients are already asking. The only open question is whether the hospital is part of the answer.
Indian hospitals may assume this is an American problem, because our patients have no portal to connect. I disagree. Our patients carry PDFs, photograph reports and share them on WhatsApp. Uploading a report to a chatbot takes less effort here than linking a chart does there, and no IT team will see it in an API log. So we may be facing the same shift with less visibility of it.
That makes the response a product and content job. Give patients reports they can understand and find later. Offer an assistant on your own channels, with a narrow, approved list of administrative tasks and a clear route to a human. And make sure your hospital is findable and correct in what AI assistants say about hospitals.
What most coverage missed
The clinical debate gets the headlines, but the commercial detail in this interview sits in scheduling. Epic reports that Ochsner raised appointment confirmation rates from 43% to 52% using conversational AI over text, and Trainor estimates online scheduling at about 30%. Then comes her idea of exposing availability to consumer AI tools.
If that happens, appointment inventory becomes a distribution feed, much as hotel rooms did for travel sites. Hospitals with clean doctor data and real slot availability could be bookable inside an AI answer. Those without it may simply not appear. Add her point on sponsorship, and the marketing questions arrive quickly: who ranks, who pays, and how is it disclosed?
What I would do
- Assume the behaviour exists: add one question to patient surveys asking whether people used AI before or after their visit.
- Make reports readable: a plain-language layer on your own channels gives patients less reason to go elsewhere first.
- Clean the inventory: doctor names, specialties, timings and slots should be structured and accurate, whichever channel reads them.
- Write the consent wording now: Trainor wants training use spelled out in the connection flow. Under DPDP, Indian hospitals should hold their own flows to the same clarity.
What to watch
- Which option Ochsner picks, and whether other systems publish their numbers.
- Whether ChatGPT Health or similar record links reach India, and in what form.
- Whether any hospital chain here offers appointments inside a consumer AI tool first.
Source: healthsystemCIO. Figures as reported at the time of writing.
Sources and further reading
- Becker’s reported (beckershospitalreview.com)
- healthsystemCIO (healthsystemcio.com)
Questions people ask
Ochsner’s CIO said about 4,500 patients had done so by 30 September 2026, as reported by healthsystemCIO.
Expanding Epic’s Emmie assistant, investing in its own internal model process, or doing neither, which the CIO warned would mean losing patients to ChatGPT.
I think so. Indian patients hold reports as PDFs and photos, and can upload them to a chatbot easily. Hospitals just see less of it.
Make reports easier to understand, keep doctor and slot data accurate, offer a narrow administrative assistant on its own channels and review consent wording against DPDP.
Read my takes first in Google Search
