The interesting difference between the UK's clinical answer engines is not response formatting; it is the architecture of trust, where each system's answers come from and what that provenance makes them good at. iatroX retrieves directly from UK guidance, NICE, CKS, SIGN, MHRA, relevant SmPCs on emc and peer-reviewed research, with source links, inside a platform that connects answers to calculators, question banks and CPD. Medwise is a UK clinical-search platform across national guidance whose enterprise proposition adds an organisation's own uploaded documents, making local policy searchable alongside the national layer. GPnotebook AI Answers synthesises primarily from its own curated archive of more than 35,000 GP-authored articles, supplemented by authoritative sources where required, linking back to the underlying articles and automatically recording small CPD activities for subscribers. All three descriptions are public and current at writing; the comparison below is architectural, with hands-on scoring an exercise every reader can run in an evening on the six questions provided.
Six test questions, and which architecture each favours
A straightforward CKS-style management question: all three serve it; the differentiator is what you land on, the national source itself, a search result set, or a curated secondary summary with the source one step further back, and preference here is genuinely a workflow taste. A product-specific medicines question: architecture decides, direct SmPC surfacing is the shape that fits, and any engine's answer should end at the exact emc document, per the source-of-record logic at /blog/iatrox-vs-emc-non-medical-prescribers. A local policy question: enterprise Medwise's uploaded-document search is the purpose-built shape where deployed, alongside dedicated local engines; national-first architectures answer with the national layer and should say so rather than improvise locality. A multidisciplinary long-term-condition question: cross-source synthesis with inspectable links does the work, the polypharmacy territory where organisation beats summarisation. A rare or emerging evidence question: research reach matters, and archives curated for common primary care will honestly thin out here, which is a scope property, not a flaw. A CPD follow-up: GPnotebook's automatic activity recording is a genuine convenience; iatroX's answer-to-learning loop, question into sources into reflection into portfolio evidence, is the deeper version of the same instinct, and the difference between logging that learning happened and structuring what was learned is the real comparison.
Scoring the architectures honestly
On source directness, iatroX's design is retrieval-first, primary guidance and product documents at answer level; GPnotebook's is curation-first, a consistent editorial layer with the primary sources behind it; Medwise's is search-first, breadth with the landing document deciding depth. On locality, enterprise Medwise wins where an organisation has deployed it, and nothing else on this page should pretend otherwise. On medicines depth, the SmPC-link habit is the test, product-level answers that end at emc. On uncertainty communication, all three should be held to the same standard this cluster holds everyone to: conflicts surfaced, not smoothed, and dates visible. On learning support, the loop is iatroX's differentiator by construction, with GPnotebook's CPD logging the pragmatic alternative. And on privacy and enterprise controls, deployment model matters more than brand, and institutional buyers should run the standard procurement questions regardless of which architecture appeals. The honest verdicts by user: a GP or ANP who values a concise, consistent curated summary and effortless CPD logging may prefer GPnotebook; an organisation whose binding need is searchable local policy should look at enterprise Medwise or a dedicated local engine; and the prescriber who wants direct source navigation, medicines-to-SmPC depth, calculators and the answer-to-learning loop in one platform is the user iatroX is built for, stated as positioning for you to test on the free layer rather than a crown we award ourselves.
Frequently asked questions
Can the three coexist in one workflow?
Easily, by layer: a curated summary for orientation, direct retrieval for source-grounded depth and medicines, local search where deployed for the operational layer; the architectures compete less than their marketing suggests.
Which is best for a prescribing student or new NMP?
The learning loop weighs heaviest exactly there: answers that end in sources, questions and reflection build the verification habit the competency framework demands, which is the configuration argued throughout this cluster.
How should I run the six-question test fairly?
Same questions, same day, notes on where each answer landed you and how many clicks the primary source took; the click-count to the exact supporting document is the single most revealing metric in the genre.
Does GPnotebook's curated archive make it less current than direct retrieval?
Different, not automatically less: curation adds an editorial layer with its own update cycle, direct retrieval inherits the source's currency; either way the visible date is the check, and the six-question test includes a recency probe for exactly this reason.
Where does emc sit relative to all three?
Above the comparison, as the shared destination: whichever engine you prefer, product-level decisions end at the exact SmPC, and the engine's quality is partly measured by how gracefully it sends you there.