Doximity is right about the direction of travel. The market is moving toward visible review — toward clinical AI where the trust architecture is transparent, the verification is attributed, and the governance is a product feature rather than a back-end afterthought.
PeerCheck gets several things right. Named physician reviewers create individual accountability — a cardiologist's name on a cardiology answer creates trust that anonymous AI output does not. Specialty-relevant review ensures domain expertise — the reviewer understands the clinical nuances that general-purpose AI evaluation cannot assess. Evidence strength and bias assessment add dimensions beyond simple accuracy — an answer can be factually correct but clinically misleading if it overstates weak evidence or ignores important caveats.
The participation scale is meaningful: 10,000+ physician reviewers, co-chaired by Eric Topol and Regina Benjamin, with PeerCheck-certified answers now visible in Doximity Ask. The model demonstrates that large-scale physician involvement in AI review is achievable — not just aspirational.
Why the UK Needs More Than Peer Review
Peer review addresses one dimension of clinical AI trust. The UK clinical AI ecosystem also needs several additional dimensions.
UK guideline alignment. Clinical answers for UK clinicians must be grounded in NICE, CKS, SmPC/eMC, MHRA drug safety updates, and SIGN — not US guidelines, US drug labels, or international recommendations that may not reflect UK prescribing practice.
UK prescribing information. Drug names, licensed indications, formulations, doses, and monitoring requirements differ between the US and UK. An answer validated by a US physician may not reflect the UK-licensed SmPC for the same medication.
MHRA regulatory awareness. UK clinical AI should be aware of MHRA Drug Safety Updates, UK-specific contraindications, and UK regulatory requirements that differ from FDA guidance.
Local pathway compatibility. UK clinical practice involves local referral routes, local formularies, and locally negotiated shared-care protocols that vary by ICB and Trust. National guidance provides the framework; local pathways provide the operational detail.
Safety-case thinking. UK medical device regulation (MHRA, UKCA) expects clinical safety cases proportionate to the tool's intended use. Post-market surveillance, hazard management, and incident reporting are regulatory expectations — not optional extras.
Post-market feedback. Real-world clinical use should generate quality improvement. A tool that does not have a mechanism for clinicians to report errors, unclear outputs, or potentially harmful content cannot systematically improve from clinical deployment.
Professional governance. UK clinical AI should be aligned with the professional accountability expectations of the GMC, GPhC, NMC, and other statutory regulators — where the professional retains responsibility for clinical decisions regardless of AI involvement.
iatroX's UK-Specific Trust Architecture
iatroX's trust architecture is designed around UK professional use. It combines curated retrieval from UK clinical sources, algorithmic fidelity checks (keeping outputs aligned with retrieved material), citation-aware synthesis (structuring answers around visible citations), conflict detection (identifying tension between sources), fail-safe behaviour (narrowing or abstaining when evidence is insufficient), and clinician feedback mechanisms.
The aim is not to remove professional judgement but to help clinicians and healthcare professionals retrieve and verify clinical knowledge faster — with transparent provenance, conservative behaviour under uncertainty, and governance proportionate to the clinical context.
iatroX is UKCA-marked and MHRA-registered. Its clinical AI standards are publicly described. The professional-facing design ensures the tool sits within the domain of professional judgement — not consumer self-diagnosis.
The Convergence
Doximity is showing the US market that physician validation matters. The UK market also needs source fidelity, regulatory literacy, professional accountability alignment, and fail-safe design. The strongest clinical AI systems will eventually combine both: human expert review where possible, and source-grounded, citation-first, fail-safe retrieval as the continuous foundation.