OpenAI has now launched four healthcare-related capabilities in 2026 across three products. According to a 2025 AMA survey of 1,183 physicians, 66% already use AI in clinical practice. The question for UK clinicians is no longer whether to use clinical AI, it is which clinical AI is appropriate for UK practice. This comparison addresses that question across every dimension that matters clinically: availability, guideline alignment, regulatory status, exam preparation, clinical calculators, citations and price.
UK availability
ChatGPT: ChatGPT Health is explicitly excluded from the UK, EEA and Switzerland, OpenAI's launch documentation confirms this. ChatGPT for Healthcare is enterprise-only, and as of 1 September 2026 gained a read-only Epic EHR connection and a nine-source Healthcare Public Data plugin, with no announced UK NHS Trust deployment of either capability. ChatGPT for Clinicians launched 22 April 2026 and appears US-first, with verification requiring an NPI, a US-specific identifier UK clinicians do not hold; eligible accounts gained the Healthcare Public Data plugin alongside the enterprise product on 1 September, without Epic access. Standard ChatGPT is available globally, but is not optimised for clinical use and lacks the safety features, citations and clinical-data integrations of the three health-specific products.
iatroX: available now, free, on web, native iOS and Android apps. UKCA-marked, MHRA-registered. No geographic restrictions for UK clinicians, no waitlist, no verification barrier.
UK guideline alignment
ChatGPT: draws from PubMed, clinical guidelines globally, and peer-reviewed literature without jurisdiction-specific weighting. US clinical guidelines are substantially over-represented in its training data, reflecting the larger volume of US medical content online. A UK GP asking what the first-line treatment for hypertension is may receive an answer aligned to ACC/AHA thresholds and drug recommendations rather than NICE NG136. Treatment thresholds, first-line drug choices and step-up protocols differ materially between US and UK guidelines, and a UK clinician asking about antibiotic prescribing may receive recommendations aligned to IDSA guidance rather than NICE antimicrobial stewardship or local Trust formularies. The 1 September Healthcare Public Data plugin does not change this picture: its nine sources, PubMed and ClinicalTrials.gov aside, are specifically US-rooted, DailyMed, RxNorm, openFDA, CMS Coverage, CMS Open Data, Medicare Care Compare and the NPI Registry among them, with no UK-guideline connector added.
This is not a ChatGPT deficiency, it is a structural consequence of using a global model for jurisdiction-specific clinical questions. The model cannot reliably determine which guideline framework applies to a patient unless explicitly specified, and even then its coverage of the latest NICE guideline version may be incomplete or outdated.
iatroX: explicitly integrates NICE guidelines, CKS summaries, peer-reviewed literature and SmPC data via emc, the reference standards UK clinicians use daily, cited in GP referral letters, hospital discharge summaries and clinical documentation. Ask iatroX answers ground in UK authoritative sources with guideline numbers, NICE NG28 for type 2 diabetes among them, citable in clinical records, referral letters and appraisal documentation.
Regulatory status
ChatGPT: not a registered medical device in any jurisdiction. No MHRA registration, no UKCA marking, no DCB 0129 clinical safety case, no post-market surveillance obligations, no systematic hazard identification or risk-control framework. The Royal College of Surgeons Bulletin, October 2025, noted that in the UK, large language models intended for medical purposes, including clinical decision support, are classified as medical devices.
iatroX: UKCA-marked Class I medical device, MHRA-registered, DCB 0129 clinical safety governance with systematic hazard identification and risk controls, post-market surveillance obligations. This is a substantive clinical safety framework requiring the manufacturer to identify clinical hazards, implement risk mitigations, maintain an ongoing clinical safety case, and monitor real-world safety through post-market surveillance. OpenAI's products have not undergone this process, including the newly Epic-connected enterprise product, whose own reported 99.1% safety figure remains a self-evaluated result rather than independent regulatory verification.
Exam preparation
ChatGPT: no exam preparation functionality. It can answer medical knowledge questions conversationally, but offers no structured Q-banks, adaptive learning algorithms, spaced repetition scheduling, mock exam modes, performance analytics, or progress tracking across topics. You cannot run a timed mock MRCP Part 1 on ChatGPT.
iatroX: 15-plus exam banks covering MRCP Part 1, MRCGP AKT, MSRA, PLAB 1, UKMLA, MRCEM, PSA and PANE, all free, plus specialist diploma banks, DRCOG, DFSRH, DGM, DipIMC, FFICM, DTM&H and GPhC CRA at £99 a year. AI-adaptive question selection driven by cumulative performance data, spaced repetition at evidence-backed intervals, mock exam mode with global countdown timers and deferred explanations, and performance analytics showing weak topics and readiness trends.
Clinical calculators
ChatGPT: can perform calculations if prompted with the right variables in free text, with no dedicated validated calculator library. There is a measurable risk of calculation errors in conversational responses, an LLM computing a CHA2DS2-VASc score or NEWS2 in free text may produce incorrect results without the structured input validation, range checking and error handling a purpose-built calculator provides, and no way to verify the correct scoring algorithm was used for a complex calculator.
iatroX: 80-plus clinical calculators with editorial content, evidence summaries, guideline references and dedicated scoring modules for NEWS2, MELD, MELD-Na, Glasgow-Blatchford, PECARN, Canadian C-Spine, YEARS PE, SOFA, Lille and APACHE II. Complex risk engines, QRISK3, SCORE2 and FRAX among them, are implemented as editorial-rich pages linking to official validated calculators, each with clinical context, when to use it, what the score means, which guideline recommends it, and what to do with the result.
Citations
ChatGPT for Healthcare (enterprise): includes citations from peer-reviewed studies with title, journal, authors and publication date, now supplemented by direct PubMed access through the 1 September Healthcare Public Data plugin. Good for research-level evidence synthesis. Standard ChatGPT may cite studies with higher hallucination risk; fabricated citations are a documented problem.
iatroX: cites UK guidelines with NICE guideline numbers, CKS topics, and SmPC data, different citation ecosystems optimised for different clinical contexts. iatroX's citations are built for the question UK clinicians ask dozens of times a week, what does NICE say about this, with verifiable guideline numbers.
Price
ChatGPT: standard ChatGPT is available on free and paid tiers, Go and Plus at £16 a month, Pro at £160 a month. ChatGPT for Clinicians appears free. ChatGPT for Healthcare is enterprise-priced with custom quotes, its 1 September Epic and Healthcare Public Data capabilities included within that enterprise pricing.
iatroX: core clinical AI, all major UK exam banks and all clinical calculators are free. Specialist diploma Q-banks are £99 a year.
Use case analysis
What's the NICE-recommended first-line for newly diagnosed type 2 diabetes? ChatGPT may give a correct answer but may cite ADA guidelines rather than NICE NG28; the specific drug, HbA1c threshold for treatment initiation, and stepped management approach differ between US and UK practice. iatroX retrieves the NICE NG28 recommendation directly, UK-specific, guideline-numbered, citable in clinical documentation.
Summarise the latest evidence on SGLT2 inhibitors in heart failure. ChatGPT is strong here, access to broad PubMed literature and strong evidence synthesis across multiple studies is where general-purpose AI excels, an advantage the direct PubMed connector strengthens further. iatroX can retrieve relevant guideline and literature evidence, though for open-ended synthesis across dozens of studies, ChatGPT has the broader reach.
I need to practise MRCP Part 1 questions. ChatGPT cannot do this: no Q-bank, no adaptive learning, no performance tracking. iatroX provides a free MRCP bank with AI-adaptive learning, spaced repetition and mock exam mode.
Calculate this patient's QRISK3 score. ChatGPT would need all 22-plus variables entered manually in free text, with no input validation and a risk of calculation error. iatroX provides a dedicated QRISK3 calculator with structured fields, editorial content and guideline context.
Help me draft a referral letter for a patient with suspected inflammatory bowel disease. ChatGPT is strong here, structuring clinical information into professional correspondence is a general-purpose AI strength, though the output should be reviewed for US guideline references before sending, a referral letter citing AHA/ACC guidance will confuse the receiving UK gastroenterologist.
The honest position
Both tools have a place. ChatGPT is a powerful general-purpose AI, now with a meaningfully broader evidence and record-retrieval stack at the enterprise level following the 1 September announcement. iatroX is a purpose-built clinical AI for UK practice. They are not interchangeable, and many UK clinicians will use both.
ChatGPT is stronger for open-ended research questions across broad literature, evidence synthesis spanning dozens of studies, writing assistance, letters, reports, audit methodology and plain-English patient explanations, non-clinical administrative tasks, and brainstorming differential diagnoses for educational purposes. iatroX is stronger for UK guideline retrieval with NICE guideline numbers, exam preparation with adaptive learning and spaced repetition, clinical calculators with validated scoring, CPD documentation, and any workflow where UK-specific accuracy and regulatory governance matter, which includes most clinical decision support during patient-facing consultations.
ChatGPT is a Swiss Army knife, now with a considerably larger toolkit at the enterprise level. iatroX is a scalpel, built specifically for UK practice. Both cut. The clinical context determines which you reach for.
Frequently asked questions
Does the 1 September Epic integration change which tool a UK clinician should use?
Not directly: the integration is available only within enterprise ChatGPT for Healthcare deployments, none announced for the NHS, and it adds no UK guideline, medicines or NHS-system grounding, leaving the core comparison in this article unchanged for individual UK clinical practice.
Could a UK NHS Trust adopt ChatGPT for Healthcare's new Epic capability?
Only where the Trust already runs Epic and would need to separately address UK GDPR, MHRA medical device classification and clinical safety governance requirements that this announcement does not itself satisfy, the same regulatory gap this comparison's regulatory-status section describes for OpenAI's products generally.
Is it ever appropriate to use both ChatGPT and iatroX in the same working day?
Yes, and many UK clinicians already do: broad literature synthesis or drafting tasks suit ChatGPT's strengths, while any UK-specific guideline, medicines or exam-preparation task suits iatroX, with the honest position section above setting out which tool fits which task.