Heidi Evidence in the UK: What the In-Session Restriction Actually Blocks

Heidi Evidence is the most strategically interesting clinical AI tool to land in 2026. It is free at the individual tier. It is citation-backed. It is partnered with NICE, BMJ Group, MIMS and HealthPathways — the four most credible UK clinical-content sources a tool in this space could pair with. We have read the output carefully. It is good.

In the UK, it will not answer from inside an active session.

That is a narrower limit than it is usually reported to be, including in earlier versions of this article, so it is worth stating precisely. Heidi documents Evidence used outside a session as "fully available in the UK and EU on all plans" — opened from the sidebar, independently of any consultation. You can use it with a patient in the room. What you cannot do is query Evidence from inside an active Heidi session, where it would read the transcript and linked patient details, and patient or session linking is unavailable for UK and EU accounts.

There is also a second limit that matters more to most readers of this site: Heidi states that NHS UK accounts are excluded from Heidi Evidence, and from free trials that include it. An account registered with an NHS UK email address stays on its current plan.

So the honest position is three-part: out-of-session Evidence works here, in-session Evidence does not, and NHS accounts are excluded from the feature altogether.

Update — 16 July 2026: Heidi Evidence formally launched on 10 July 2026, alongside a $65M raise, the acquisition of AutoMedica — a UK clinical AI firm with a place in the MHRA's AI Airlock pilot — and a new communications product, Heidi Comms. None of it changes the UK position described below: Heidi's own help pages still document Evidence as available to UK and EU clinicians only outside live sessions (re-verified 15 July 2026). If anything, the AutoMedica acquisition is the clearest signal yet that Heidi intends to resolve the UK regulatory question properly. We will update this post when the in-session restriction changes.

What Heidi Evidence actually does

Heidi Evidence is a clinical decision support tool. You ask a clinical question — the kind of question you might have asked a colleague on the way to coffee — and it returns a synthesised answer with citations you can click through to.

The synthesis is grounded in named, recognisable UK sources. The NICE partnership means guidelines come from where they should come from. The BMJ Group content gives the answers a familiar editorial register. MIMS gives prescribing detail. HealthPathways adds local-pathway awareness for the regions where it is deployed.

Two things make Heidi Evidence stand out from earlier clinical AI tools.

The first is that the citations are honest. Click any reference and you land on the source — NICE guideline, BMJ article, MIMS entry — exactly as quoted. This is not always true of clinical AI tools, and it has been a real problem with some of the larger-funded US entries.

The second is that there is no advertising and no sponsored content. The product is free at the individual tier and the funding model does not depend on shaping which answers surface.

For UK clinicians, the catch is how you invoke it, not when. From the sidebar, it is available. From inside a session, it is not.

What the in-session restriction actually looks like

The restriction is on the mode of use, not on access or timing. A UK or EU clinician can hold a Heidi Evidence account and open the sidebar tool whenever they like, including mid-clinic. What they cannot do is run the query from inside an active session, or attach a patient record or live session to an Evidence conversation.

In practice that means the tool works, but it works blind. "I have ten minutes between patients and want to check the current line on a drug interaction" is fully supported. So, in fact, is "the patient is in front of me and I need to look something up" — provided you ask it as a general question in the sidebar. What is not supported is the version where Evidence can see the consultation: the transcript, the linked patient details, the context that would let it answer about this patient rather than the topic.

⚠️ NHS UK accounts are a separate matter. Heidi excludes them from Evidence entirely, so for a clinician whose account is an NHS UK email address, none of the above applies — the feature is not available to them at all.

Heidi Scribe — the AI clinical scribe that drafts consultation notes — is a different story. Heidi Scribe is separately MHRA-cleared and remains available to UK clinicians, including NHS staff, in session. If you have been using Heidi Scribe, that is unaffected. The restriction is specific to Heidi Evidence's in-session use.

Why the restriction is in place

Heidi now publishes its rationale, and it is more specific than the speculation this article previously carried. Heidi operates Evidence in the UK and EU as a non-medical-device product, and states that in-session Evidence carrying patient context "would constitute clinical decision support, which requires medical device classification under UK MDR 2002 (as amended)."

That is a precise regulatory line rather than general caution: it is patient context inside a session, not AI-synthesised clinical answers as such, that would move the product into device territory. The NICE, BMJ Group, MIMS and HealthPathways partnerships are content arrangements — they describe what the model is grounded in, not how the model is classified for live clinical use.

The wider UK regulatory framework for AI in healthcare is still mid-flight. The Medical Devices (Amendment) Regulations consultation closed on 19 June 2026. The International Reliance Framework is planned for Autumn 2026. The National Commission on AI in Healthcare report is due later in the year. Heidi's position sits inside that unsettled rulebook, but the company's own stated reason is the narrower device-classification one above, not general uncertainty.

The July 2026 acquisition of AutoMedica points the same way. AutoMedica holds a place in the MHRA's AI Airlock pilot — the regulator's sandbox for AI medical devices — which gives Heidi a direct working relationship with the process that will decide how tools like Evidence are classified for live clinical use. That is not a clearance, but it is the posture of a company planning to lift the restriction properly rather than sidestep it.

This sits in the same regulatory current that took OpenEvidence out of the UK and EU on 28 April 2026 and has kept ChatGPT for Clinicians US-first since 23 April. The difference is one of degree: those two stepped back entirely; Heidi Evidence stayed, with a limit on how it can be used.

The pattern UK clinicians should be aware of

In the space of six weeks, three clinical AI tools in the same category changed their UK footing:

• 23 April 2026: ChatGPT for Clinicians launches. UK and EEA excluded at the verification step.
• 28 April 2026: OpenEvidence withdraws from the UK and EU.
• 2026: Heidi Evidence operates in the UK and EU, but only outside an active session — and NHS UK accounts are excluded from the feature entirely.

All three offered the same broad use case — clinical Q&A grounded in cited sources. Two are now unavailable to UK clinicians; the third is available, but not in the mode that would let it see the patient in front of you, and not at all on an NHS account.

We have written separately about the OpenEvidence withdrawal and the ChatGPT for Clinicians exclusion. Heidi Evidence is the third chapter in the same regulatory story. If you are choosing between the two directly, see Heidi Evidence vs OpenEvidence.

What is striking is that Heidi Evidence — unlike the other two — is the one product that did everything right on the UK content side. NICE partnership. BMJ Group partnership. UK-relevant prescribing data. And it still holds back the patient-context mode. That is device regulation talking, not the product.

What UK clinicians can use right now

For clinical Q&A that needs the patient's context — inside the consultation, aware of the case: iatroX is the UK-based option. It holds MHRA registration as a Class I device, carries UKCA marking, and is built around UK guideline retrieval (NICE, CKS, SmPC) with calculators and Q-banks attached. It is not a perfect substitute for Heidi Evidence's model quality, but it is the most credible UK-regulatory-compliant in-session tool today.

For general evidence questions: Heidi Evidence's sidebar tool is available and worth using — between patients, before clinic, after a consultation, or during one, so long as the question is a general one rather than about the patient in front of you. Not available on an NHS UK account.

For AI clinical documentation: Heidi Scribe remains available and MHRA-cleared, in session included.

For staying current with the medical literature — knowing what was published in your specialty each week — that is a different job altogether. Q&A tools answer questions you have already thought to ask. They cannot tell you about the practice-changing trial that landed last Tuesday. That is the gap The Monday Clinical Brief is built to close.

We have written more on how Heidi Evidence and a weekly digest fit together — push and pull, current awareness and point-of-care.

A short note on Claude

Heidi Evidence is powered in part by Anthropic Claude. So is The Monday Clinical Brief — we use Claude to summarise the weekly journal output for each of our 31 specialties. We mention this only because it is worth knowing: the underlying model choice is the same. The product around the model is what differs. Heidi Evidence wraps the model in a query interface and clinical-source partnerships. MCB wraps the model in a weekly editorial cadence and a UK clinician's perspective.

What we offer

The Monday Clinical Brief sends one email every Monday morning. In that email are the most important peer-reviewed papers published in your specialty over the previous week — plain-language summaries, links to the originals, and the UK context (NICE, Royal College, NHS pathway) where it matters.

We cover 31 specialties. We are reader-paid at £20 a year. No advertising, no sponsored papers, no regulatory uncertainty — we are not a medical device and we do not claim to be one. We are a weekly reading habit you can log for CPD.

Subscribe to The Monday Clinical Brief — and know what was published last week before any of these AI tools could have told you about it.

The bottom line

Heidi Evidence is excellent, and in the UK and EU the sidebar tool is available to use whenever you like — including with a patient in front of you. What it will not do here is answer from inside an active session with the consultation's context attached, which is the version that would know about this patient rather than the topic. And if your account uses an NHS UK email, Heidi excludes you from Evidence altogether.

The in-session limit may ease as the UK regulatory framework for clinical AI settles, though Heidi's stated reason — device classification under UK MDR 2002 — is a specific bar to clear rather than a general wait.

In the meantime: iatroX for in-session UK clinical Q&A, Heidi Evidence for out-of-session questions, Heidi Scribe for documentation if you want it, and MCB for keeping current with the literature itself. Each tool does one job well. Together they cover most of the ground.

We will update this post as Heidi Evidence's UK position changes. Any change will likely follow the UK regulatory process now in train — the Medical Devices (Amendment) Regulations consultation closed on 19 June 2026, and Heidi's acquisition of AutoMedica gives it MHRA AI Airlock access and a seat close to that process. Until something lands, the in-session restriction and the NHS account exclusion both hold.

Heidi's UK and EU position was re-checked against the company's own help documentation on 30 August 2026. An earlier version of this article said Heidi Evidence could not be used during a consultation; that was too broad, and the position above is Heidi's own.