AI in medicine, explained for everyoneGeneral information only, not medical advice · Our standards
How Far

Diagnosis (general) · AI doctors and chatbots

Microsoft's MAI-DxO: AI 'panel of doctors' on hard cases

How it was checked
AI-assisted source check. Not clinically reviewed. ()
Last updated
Next review
Page ID
TR-009

Our editorial standards · 2 references · Corrections

General information only, not medical advice. Talk to your doctor about your situation. Read the disclaimer.

In June 2025 Microsoft reported MAI-DxO, an AI system that works through difficult published case studies step by step. It scored much higher than doctors in a test Microsoft designed. It's a preprint, not peer-reviewed, and was tested on case records, not patients.

How far away is this?

A non-peer-reviewed preprint on case records. We rate it Early research because there's published data, but it's the weakest kind.

Where is it approved?

RegionRegulatorStatusNotes
United StatesFDANot approvedResearch demonstration; not a product.
European UnionCE mark / EMANot approvedResearch demonstration; not a product.
United KingdomMHRANot approvedResearch demonstration; not a product.
AustraliaTGANot approvedResearch demonstration; not a product.

A research demonstration, not a product. "Unknown" means we haven't verified it, not that it's unavailable. Approval in one country isn't approval in another.

What is it?

MAI-DxO sets several AI 'roles' to debate a case, order tests and reach a diagnosis, while tracking the cost of tests.

What's the evidence?

  1. Microsoft's paper, written by its own researchers, built a benchmark from 304 New England Journal of Medicine case records and reported MAI-DxO reaching up to 85.5% accuracy in its best setup. These are selected teaching cases, chosen because they're unusually hard, so they aren't typical of everyday medicine.[1]

    Evidence stage: Early research · #c1

  2. The 21 physicians in the comparison (from the US and UK) averaged 20% accuracy, but they worked without colleagues, textbooks or other usual resources, which isn't how real practice works.[1]

    Evidence stage: Early research · #c2

  3. Microsoft describes it as research, not a clinical tool.[2]

    Evidence stage: Announced only · #c3

What does it mean for me?

Headlines about AI 'beating doctors' 4 to 1 come from this kind of test. It's interesting, but it doesn't tell us how it would do with you.

What could I ask my doctor?

  • How do you use AI tools, if at all, in your practice?
  • If my diagnosis is unclear, what are the next steps?

These are general starting points, not a recommendation to choose any particular test or treatment.

Change history

  • : First published.
  • : Updated after independent fact-check and clinical review
  • : Review label corrected: this page was not clinically reviewed. Earlier labels saying 'independent fact-check' and 'clinical review' were wrong; checks so far were AI-assisted source checks only.

General information only, not medical advice. Talk to your doctor about your situation. AI-assisted source check. Not clinically reviewed.

References

  1. Nori et al., arXiv preprint (2025). arxiv.org (2025). Preprint (not peer-reviewed). https://arxiv.org/abs/2506.22405
  2. Microsoft AI: The path to medical superintelligence. microsoft.ai. Company document (claim). https://microsoft.ai/story/the-path-to-medical-superintelligence/

Publisher shown as the website hosting each source. Dates are given where the source label includes them.