Is there a medical AI like ChatGPT?

Yes. The medical AI tools that work like ChatGPT in 2026, for patients and clinicians, how they differ from a regular chatbot, and how to use them safely.

Woman typing a health question on her phone at the kitchen table in the evening

Yes. There are now several medical AI tools that work a lot like ChatGPT: you type or speak a question and get an answer in plain language. Which one is right depends on who you are. Some are built for patients, some only for verified clinicians, and some are still research projects. Here is the landscape as of October 2026.

The short answer

For patients: ChatGPT itself now has a dedicated health space, and Claude, Google and Perplexity have added health features. Most of the features that connect to your medical records are available only in the United States so far.
For clinicians: tools like OpenEvidence, ChatGPT for Clinicians and Doximity GPT answer clinical questions with citations to medical literature, and are free for verified clinicians. Clinic-side tools like Athel go further and work with patients directly before the visit, with the physician reviewing everything.
In research: systems like Google's AMIE and Microsoft's MAI-DxO show where things are heading, but they are not available for clinical use.

Medical AI for patients

ChatGPT Health (OpenAI)

OpenAI announced ChatGPT Health in January 2026 and, in July 2026, opened it to logged-in US adults on its free and paid plans. It is a separate health space inside ChatGPT that can connect to Apple Health and to records from US hospital systems. OpenAI says connected records are not used to train its models, and that disconnecting deletes the synced data within 30 days. It also states plainly that the tool does not replace qualified medical professionals. It is not available in Canada.

Claude (Anthropic)

In the US, Claude Pro and Max subscribers can choose to connect lab results and health records, as well as Apple Health and Android Health Connect. Anthropic positions it for summarizing your history, explaining results in plain language and preparing questions for appointments, and says the data is not used for training.

Google Health Coach

Google renamed the Fitbit app to Google Health in May 2026 and added a Gemini-powered coach. It comes with Google Health Premium at US$9.99 a month and is focused on fitness, sleep and general wellness. Google describes it as for general wellness and information only, not for diagnosing or treating any condition.

Perplexity Health

Perplexity added health features for its Pro and Max subscribers in the US in 2026, with connections to Apple Health, wearables and medical records.

Symptom checkers

Older, more focused tools still matter. Ada, a symptom checker headquartered in Berlin with an office in Toronto, is certified as a Class IIa medical device in the European Union. That is a meaningful difference from general chatbots, which are not regulated as medical devices.

Medical AI for clinicians

OpenEvidence answers clinical questions using medical literature and is free for verified clinicians, supported by advertising. The company reported about 18 million clinical consultations a month by late 2025, and says verified Canadian physicians can now sign up.
ChatGPT for Clinicians launched in April 2026 and is free for verified US physicians, nurse practitioners, physician assistants and pharmacists. It offers cited clinical search and help with tasks like referral letters, and a business associate agreement for HIPAA use is available to eligible accounts.
Doximity GPT is free for clinicians on Doximity, a network mainly for US-licensed professionals. It answers with citations and helps draft paperwork such as prior authorizations.
Athel is an AI co-clinician built in Canada for clinics. Before the visit, it interviews patients by voice, web chat or video and hands the physician a structured, prioritized history to review, and its AI receptionist answers the clinic phone 24/7. It is HIPAA and PHIPA compliant, keeps a clinician in the loop on every summary, and verified physicians can try it free for 30 days.
UpToDate Expert AI from Wolters Kluwer answers only from UpToDate's expert-written content and is sold to health organizations rather than offered free.

Medical AI still in research

Google's AMIE is a research system designed to take a patient's history through conversation. In a study published in The Lancet in October 2026, 98 patients chatted with AMIE before their appointment, and its list of possible diagnoses included the final diagnosis 90% of the time. Microsoft's MAI-DxO solved 85.5% of 304 difficult published cases in a 2025 research demonstration. Neither is approved for clinical use, and both companies say so.

How is medical AI different from regular ChatGPT?

The differences that matter are not about how the chat looks. They are about four things:

Sources. Clinician tools such as OpenEvidence and UpToDate Expert AI answer from vetted medical literature and show citations. General chatbots draw on everything they were trained on.
Regulation. A few tools, like Ada, are certified medical devices. General chatbots are not regulated as medical devices, which is one reason the patient safety group ECRI named misuse of AI chatbots the top health technology hazard for 2026.
Privacy. US health privacy law (HIPAA) covers healthcare providers, not chatbot companies. What you upload to a consumer chatbot is protected by that company's own privacy policy. In Canada, laws like Ontario's PHIPA govern how clinics and their vendors handle health information.
Purpose. Some tools are built to inform you, some to support a clinician, and some to work inside a clinic's workflow. Using a tool for a job it was not built for is where most problems start.

Do these tools actually give good answers?

On tests, often yes. In practice, it depends heavily on how they are used. In a randomized study from the University of Oxford, published in Nature Medicine in February 2026, AI models alone identified the right condition about 95% of the time. When members of the public used the same models, they named a relevant condition in fewer than 35% of cases, no better than people who did not use AI.

A 2025 Mount Sinai study found that chatbots given a case containing one invented medical detail went along with the false detail and expanded on it in 50% to 83% of cases, depending on the model. Adding a short warning to the prompt reduced this, but did not eliminate it.

How to use a medical AI safely

Use it to prepare, not to decide. It is good for understanding terms, organizing your symptoms and writing down questions for your appointment.
Give it full context, and be careful about what personal information you share.
Ask for sources, and be suspicious of confident answers without them.
Never delay urgent care because a chatbot reassured you. If something feels serious, contact a clinician or emergency services.

Where AI fits on the clinic side

There is a third kind of medical AI that patients meet without searching for it: tools their clinic uses. Clinic-side systems like Athel, an AI co-clinician, interview patients before their visit and hand the physician a structured summary to review. The difference from a consumer chatbot is that a clinician is always in the loop and the information goes to the people treating you.