How we evaluate AI solutions
Vendor websites tell you what an AI tool promises. Our profiles tell you what is actually known: what the tool does, how it is regulated, what the independent evidence shows, and where its limits lie. This page explains the rules every profile follows.
What we include
AI4Hospitals lists solutions that are in real use, not research prototypes or demos.
- The solution is deployed in at least one hospital or healthcare organisation, in clinical or operational use.
- Where the tool qualifies as a medical device, its regulatory status (EU MDR / CE marking, FDA) is reported. Tools that are not medical devices (for example operational or administrative tools) can be listed, and are clearly labelled as such.
- Promising research projects without real-world deployment are not listed until they reach that stage.
What every profile documents
Each solution profile follows the same seven-part structure, so tools can be compared on equal terms.
- 1
Precise indication
The exact use case, the population covered (adult, paediatric) and the data or imaging modality involved, not a generic label.
- 2
Regulatory status
Which clearances or markings exist, for which indication, and when. If we cannot confirm a status, we say so explicitly rather than assume it.
- 3
Clinical evidence, on two levels
We always separate vendor-supported studies from independent or systematic reviews. Vendor-backed evidence alone is never presented as sufficient.
- 4
Real-world deployment
A verifiable figure (sites, hospitals) rather than "used worldwide", with a note on whether the number comes from the vendor or an independent source.
- 5
Limits and uncertainties
What remains unproven, mixed or at risk of bias. This is the part most often missing from marketing, and the most important for decision-makers.
- 6
Role in the clinical workflow
Decision support or autonomous use, and who remains responsible for the final decision.
- 7
Dated, linked sources
Every factual claim is backed by a source with a date and a direct link, so readers can check it themselves.
How we grade evidence
Each profile shows the highest level of evidence we found, so you can see at a glance how solid the claims are.
| Level | What it means |
|---|---|
| A | Independent systematic review or meta-analysis, or independent prospective multicentre study. |
| B | At least one independent peer-reviewed study (not funded or co-authored by the vendor). |
| C | Peer-reviewed study supported, funded or co-authored by the vendor. |
| D | Vendor claims only: white papers, press releases, case studies without peer review. |
Independence and corrections
Our independence
- AI4Hospitals is non-commercial. Vendors cannot pay to be listed, featured or ranked.
- Vendors may send us information, but it is treated like any other source and labelled as vendor-provided.
- Contributors declare any link with a vendor they write about.
Updates and corrections
- Each profile shows the date it was last reviewed.
- Profiles are re-checked at least once a year, or sooner when a new clearance or major study is published.
- Spotted an error or a missing source? Tell us and we will review it; corrections are dated on the profile.
What our profiles are not. They are not medical advice, not an endorsement of any product, and not a substitute for your own due diligence.
Before procuring or deploying an AI solution, hospitals should carry out their own assessment, including regulatory compliance, data protection (GDPR), cybersecurity, integration with existing systems and local clinical validation.
Methodology last updated: September 2026.