
AI tools are spreading quickly through European health systems, but the evidence on whether they improve patient outcomes has not kept pace. Nearly three-quarters of EU Member States, 74 percent, use AI in diagnostics, according to a WHO/Europe report published in April, while most clinical trials of AI remain small and focused on technical performance rather than health results.
The gap matters to health systems, regulators and technology buyers because spending decisions are being made faster than the clinical evidence base is growing. An analysis published Oct. 15 by Innovation News Network argued that AI in healthcare adoption should be guided by evidence of benefit to patients, not by accuracy scores or deployment counts alone.
[TABLE OF CONTENTS]
How Widely AI in Healthcare Is Used in the EU
The WHO/Europe report, the first snapshot of AI in health care across all 27 EU Member States, drew on data collected from June 2024 to March 2025. It was produced under a funding agreement with the European Commission. Key findings include:
- 74 percent of EU countries use AI in diagnostics, including medical imaging, disease detection and clinical decision-making.
- 63 percent use chatbots to support patient engagement.
- Nearly half have created dedicated professional roles for AI and data science in health.
- 81 percent actively involve stakeholders in shaping AI governance in health.
All 27 countries cited improved patient care as a reason for developing AI, according to WHO/Europe. The figures describe national-level use, however, and do not show how many hospitals or patients are actually reached by the tools.
Approvals Are Not the Same as Outcomes
The pace of regulatory authorization tells a similar story. The U.S. Food and Drug Administration lists more than 1,600 AI-enabled medical devices authorized as of September 2026, according to the Innovation News Network analysis. The FDA’s list is not comprehensive, and an authorization shows that a device met regulatory standards, not that it has been widely deployed or shown to improve survival.
Clinical trial evidence is thinner. A 2024 scoping review identified 86 randomized trials of AI in clinical practice from 2018 through late 2023, the analysis noted. Most reported positive primary endpoints, but more than half measured diagnostic yield or performance, and most were conducted at a single center. That leaves open whether results hold up across different hospitals, staffing levels and patient populations.
The author argued that a better prediction can improve a dashboard long before it changes a patient’s health, and that evaluations should follow the full care pathway, from the AI output to the clinical decision and the eventual outcome.
Europe’s Next Wave of Investment
The European Union is moving from pilots toward organizing services around AI. The European Commission’s network of AI-powered advanced screening centres held its first meeting May 11, and a Digital Europe Programme call that closed Oct. 1 offered 9 million euros for two large-scale pilots of AI in medical imaging, according to the analysis. Those are funding commitments, not completed deployments.
The EU is also implementing the AI Act, which WHO/Europe described as the world’s first comprehensive legal framework specifically regulating AI. Under current arrangements, clinicians remain legally and ethically responsible for decisions supported by AI tools, even when they may not fully understand how those tools work, the WHO report noted.
Frameworks for Better Evidence
Several frameworks already exist to raise evidence standards. The DECIDE-AI reporting guideline, published in The BMJ in 2022, sets out what early-stage clinical evaluations of AI decision support should report. In England, NICE’s Evidence Standards Framework for digital health technologies describes the evidence buyers should expect, though NICE notes that meeting the framework does not amount to NICE endorsement or regulatory approval.
WHO/Europe recommended that governments strengthen workforce training on AI fundamentals, ethics and data governance; involve health professionals, patients and the public in AI policy; and establish centres of excellence to test technologies and develop common standards. It warned that systems built without public input may face resistance and could widen health inequities.
What Comes Next
For health system leaders, the practical lesson is to define what each AI investment is meant to prove before buying it, whether that is faster diagnosis, fewer admissions or lower costs, and to measure that outcome after deployment. Comparing results across hospitals with different resources would show whether gains travel beyond well-equipped academic centers.
As screening networks and imaging pilots expand, the evidence they generate will determine whether AI in healthcare delivers on its promise for patients or remains mainly a story of adoption metrics.
For more healthcare industry updates, insights and news, visit DistilINFO. Click here to subscribe to stay informed.
