BMShop Products
← All articles
Microsoft AI DxO Diagnoses More Accurately Than Doctors
AI News

Microsoft AI DxO Diagnoses More Accurately Than Doctors

Microsoft's MAI-DxO system achieved 83% diagnostic accuracy versus 36% for generalist physicians in a study published in Science on October 22, 2026.

Ammar Rahimi · 23 October 2026 · 4 min read

Microsoft announced a new medical AI diagnostic system that recorded more than double the accuracy of generalist physicians, according to a study published in the journal Science on October 22, 2026. The system, called MAI-DxO, achieved 83.2% accuracy on complex cases, compared with 36.1% for generalist physicians using reference materials.

1. What is MAI-DxO

MAI-DxO (Microsoft AI Diagnostic Orchestrator) is a system that combines multiple AI agents to work like a panel of doctors. Each agent plays a different role: some ask the patient questions, some suggest tests, and some debate the diagnosis.

The system does not receive a complete list of symptoms at once. Instead, it starts with the patient's chief complaint, then determines which follow-up questions to ask and which tests to order before arriving at a diagnosis. This approach mirrors how doctors actually work in a clinic.

Microsoft tested MAI-DxO on 304 cases from the New England Journal of Medicine and 325 emergency department admissions from the MIMIC-IV-ED dataset. All 15 study authors are Microsoft employees who hold equity in the company, and Microsoft funded the research.

2. The accuracy numbers recorded

On complex cases from NEJM, MAI-DxO achieved 83.2% accuracy. Generalist physicians using reference materials scored 36.1%. Without those resources, physician accuracy dropped to 22.3%, making the AI system nearly four times more accurate.

On emergency department cases, the system reached 80.3% accuracy compared with 46.8% for physicians. The margin here was smaller than on complex cases. Microsoft's official blog states the system is 1.7 times more accurate on emergency cases and more than twice as accurate on complex cases.

Becker's Hospital Review reported different cost figures. With no budget limits, MAI-DxO averaged $5,431 per complex case. With budget constraints, it hit 80.9% accuracy at $2,241. Physicians using search and reference tools averaged $4,725 per case.

3. The AI models underneath

MAI-DxO is not a single AI model. It is an orchestration layer that coordinates multiple models from different companies. Microsoft tested the system with models from OpenAI, Google, Anthropic, Meta, and DeepSeek.

The best results came with GPT-5 from OpenAI. With the newer GPT-5.5, DxO raised accuracy from 72% to nearly 87% on complex cases at a slightly lower cost. However, on emergency department cases, the improvement was negligible: both versions scored around 85%.

TechCrunch reported that much of the improvement came from more powerful underlying AI models, not from Microsoft's orchestration system itself. DxO added an average of 7.9 percentage points on complex cases and 4.4 points on emergency cases.

4. Study limitations and clinical trials

The study authors themselves wrote that the comparisons "do not establish superiority over specialists, multidisciplinary teams, or real-world clinical practice." The most complex cases are usually handled by specialists or physician teams, not individual generalists.

The system was also tested in a text-only environment, not with physical patient examinations. The reported costs are estimates based on simulated cases, not actual medical expenses.

Microsoft will begin clinical trials at Beth Israel Deaconess Medical Center in Boston in the coming months. Patients will interact directly with the system, and performance will be measured against physicians' independent assessments.

5. The broader context

Microsoft is not the first company to claim AI can match or beat doctors. Google with its AMIE system and OpenAI with its o1-preview model previously reported similar results on simulated tasks.

However, a 2025 study cited by Gizmodo found that doctors who relied on AI to identify potential cancer risks during colonoscopies performed worse after the technology was taken away. This highlights the risk of over-reliance on AI in medical practice.

Mustafa Suleyman, CEO of Microsoft AI, described the results as a step toward "medical superintelligence." He told The Verge that Microsoft would build something "indistinguishable from human performance in the next couple of years."

What it means for you

AI diagnosis is maturing, but remains in the clinical trial stage. Here is what you need to know:

  1. Not a doctor replacement: The system was tested against generalist physicians, not specialists. It has not been compared with the multidisciplinary teams that usually handle complex cases.
  2. Clinical trials are ongoing: Microsoft will only begin testing at a Boston hospital in the coming months. Real-world clinical results may differ.
  3. Cost figures vary: Becker's Hospital Review reported MAI-DxO spending $5,431 without budget limits, while Microsoft's blog claims half or a third of physician costs. These figures are based on simulations, not actual medical expenses.
  4. Conflict of interest exists: All study authors are Microsoft employees holding equity in the company. Microsoft funded the research itself.
  5. Opportunity for Malaysia: AI diagnostic systems could help reduce the burden on doctors in public clinics if proven safe in clinical trials. Malaysia is investing in AI through AI Malaysia Berhad and Budget 2027.

Sources