As clinical AI matures, the interesting competitive question is shifting away from raw capability, can the system answer the question at all, towards trust, can a clinician actually rely on the answer it gives. Four platforms illustrate four genuinely different models for building that trust, each addressing a related but distinct part of the underlying problem.
The four competing trust models
OpenEvidence's model centres on automated evidence-strength grading, addressed throughout this series via EvidenceGrade. Heidi Evidence's model centres on regionally oriented citations integrated directly into clinical workflow, prioritising relevance to where a clinician is actually working. Doximity Ask's model centres on literature access combined with physician-review mechanisms, drawing on a large practising-physician community for a further layer of scrutiny. And iatroX's model centres on UK guideline grounding combined with evidence-hierarchy prioritisation, addressed throughout this series.
What each model is actually trying to solve
Weak evidence, treated as though it were strong, is the problem EvidenceGrade addresses most directly. Hallucinated or unsupported claims, citations that do not genuinely exist or do not genuinely support what is attributed to them, is a distinct problem regionally oriented and physician-reviewed models are particularly well positioned to catch. Lack of local relevance, evidence that is strong but does not map onto a clinician's own healthcare system, is the problem iatroX's UK-specific grounding is built around. And lack of clinician oversight, an answer with no human check anywhere in the loop, is what physician-review mechanisms specifically add.
These dimensions are complementary, not interchangeable
None of these four models fully substitutes for another. A platform that grades evidence strength well can still present hallucinated citations if it lacks strong verification. A platform with strong physician review can still surface evidence that is technically accurate but inapplicable to a different healthcare system. Genuine trust requires several of these dimensions addressed together, not excellence in just one.
Four questions worth asking of any clinical AI platform
Are the sources authentic, genuinely existing and genuinely saying what is attributed to them. Are the sources methodologically strong, by the standards covered throughout this cluster. Does the synthesis accurately represent what those sources actually say, without subtly overstating certainty or overgeneralising findings. And is the answer applicable to the clinician's own jurisdiction and healthcare system specifically.
Where iatroX sits within this landscape
iatroX is founded by a practising UK clinician, focused specifically on NHS and UK clinical application, favouring authoritative guidelines and higher-order evidence together, and offering integrated clinical reference, calculators, learning and exam preparation within a single platform built around UK clinical practice.
Being honest about what iatroX does not currently claim
iatroX does not currently offer a direct, feature-for-feature equivalent to EvidenceGrade's automated, real-time, claim-level letter grading. This is worth stating plainly rather than implying otherwise, since the two platforms have made different, deliberate design choices about where to invest first.
The genuine next competitive frontier
The more interesting frontier for clinical AI generally is not simply accumulating more citations per answer. It is clearer provenance, clearer hierarchy, clearer applicability, and honest visibility of remaining uncertainty, ideally addressed together rather than any single dimension standing in for the whole problem.