Someone funds a free medical AI service even when the individual clinician pays nothing. The money may come from an institution, philanthropy, sponsorship, another product or a combination of sources. The useful question is not whether free access is inherently suspicious, but whether the funding and responsibilities support a dependable service.
This article presents a general economic and continuity framework. It does not reconstruct the undisclosed economics of the Anthropic and OpenEvidence collaboration or attribute an unconfirmed funding model to either company. The September 2026 partnership reporting establishes the news context, not a complete financial account.
This article is published by iatroX and includes its free and paid services among the examples. For clinicians and institutions, the important distinction is between the price of entry and the cost and durability of the service behind it.
Free is a price, not a cost structure
An individual price answers a narrow question: what must this user pay for the stated entitlement? It does not establish the supplier's operating cost, the organisation's implementation effort or the time a clinician spends checking outputs.
A free service can be sustainable. A paid service can be fragile. Neither conclusion follows from the price alone. The relevant evidence concerns how costs are funded, which obligations continue and what happens when an assumption changes.
The same distinction applies to a donation. Harvard's Better Evidence programme, as described on 23 September 2026, provides donated UpToDate access through an eligibility and renewal process. The recipient's price and the arrangements enabling that access are different layers of the proposition.
This is not a reason to discount free services. It is a reason to evaluate the full arrangement rather than use "free" as either a quality endorsement or an accusation.
The costs that do not disappear
Computing is one component: generating responses, retrieving information, storing authorised material and operating the application. Its importance depends on the architecture and usage, neither of which should be guessed from a product demonstration.
Evidence access is another. Some sources are openly accessible; others have contractual conditions. A service may also need processes for indexing, version management and withdrawal of material that is no longer appropriate to use.
Clinical and operational work continues around the software. Reviewers may need to assess content, investigate errors, test changes and maintain locally relevant pathways. Users may require onboarding, support and clear explanations of the service's boundaries.
Monitoring and maintenance are not optional merely because the application is already built. A change in a source, integration or model can alter behaviour. A responsible service needs a way to identify and respond to that change.
These are general categories for analysis, not estimates of any supplier's expenditure. Without supplier-specific data, they identify what needs funding rather than establish the size or allocation of a particular company's costs.
Measure cost per useful task, not only per generated answer
A low computing cost per response can coexist with a high cost per resolved question. An answer that needs extensive correction, another search or a support request may shift work to the user rather than eliminate it.
A proposed service assessment could therefore consider the resources used to reach an appropriately checked result. That includes the application, the clinician's review effort and any follow-up needed because the answer was incomplete.
In a fictional teaching hospital, one evidence tool produces a response quickly but repeatedly omits the local pathway. Another takes longer to display text but makes the relevant source easier to inspect. The first is not necessarily cheaper in the sense that matters to the service, even if its generation cost is lower.
No performance figures are assigned to this example. It illustrates why the denominator matters. "Per response", "per active user" and "per completed task" describe different economic questions and should not be mixed in a comparison.
Several funding models can support free individual access
The following table describes possible arrangements. It does not identify the model used by the Anthropic and OpenEvidence initiative.
| Possible funding route | Potential strength | A tension or continuity question to examine |
|---|---|---|
| Institutional purchasing | The organisation funds access and may support implementation | What happens when a contract ends or the clinician changes employer? |
| Philanthropic support | Funding can target users underserved by commercial demand | How are access and maintenance sustained beyond the funding period? |
| Commercial sponsorship | A sponsor can cover costs without charging the user | How are sponsorship and influence on content or presentation disclosed and managed? |
| Cross-subsidy from another service | Revenue elsewhere can support a free reference function | Which entitlements remain free and how are changes communicated? |
| Paid adjacent services | Users pay for distinct education, integration or organisational functions | Is the boundary clear, without making the free offer misleading? |
| Shared infrastructure or partnership support | Participants contribute complementary capabilities or resources | Who remains responsible for support, source maintenance and continuity? |
Each route can be implemented well or poorly. Institutional purchasing does not automatically guarantee clinical independence, and philanthropic support does not automatically guarantee permanence. Sponsorship is not, by itself, evidence that an answer is influenced; the relevant question is the documented arrangement and its safeguards.
There is no basis here to allege advertising manipulation, data sales or hidden motives by a particular company. Such claims require specific evidence, not inference from the absence of an individual subscription charge.
Separate access conditions from editorial independence
A clear offer should state what the user receives, what limits apply and which services require payment. A clear evidence product should also explain how sources and recommendations are selected.
Those are related but separate forms of transparency. A service may have simple pricing and still provide little information about its evidence process. Another may describe careful source handling while leaving the boundaries of a free account unclear.
A useful review therefore asks both questions. Can the user understand the entitlement without guessing? Can the user inspect the source and distinguish a recommendation from commercial or organisational messaging?
For any funding route, the supplier should be able to explain how corrections are handled and whether financial arrangements affect which material is displayed. That is a request for transparency, not a presumption that an inappropriate influence exists.
Local adaptation creates continuing obligations
A service that incorporates local guidance has a maintenance problem as well as a launch problem. Someone must know which document is current, who approved it and what happens when local practice changes.
The organisation may own its guidance while the supplier operates the retrieval system. Those responsibilities should be distinguished. A commercial or funding change should not leave the organisation uncertain about which documents it can retrieve, which configurations it can retain or who will remove a superseded pathway.
A practical agreement can make the scope explicit: the local content supplied, the permitted use, responsibility for updates, support arrangements and the process when access ends. The exact terms depend on the arrangement; this article does not supply legal advice or assume a particular contract.
For users, the visible consequence is simple. A free answer should not continue to present abandoned local content as current merely because the application remains online.
A free-access, sustainable-service checklist
A useful assessment can be organised around a small set of direct questions rather than a speculative financial model.
| Question | Evidence worth requesting |
|---|---|
| What exactly is free? | Current entitlement, limits, eligibility and any separate paid functions |
| Who maintains the clinical sources? | Named responsibilities and an update or correction process |
| What local implementation work is required? | A defined account of onboarding, configuration and ongoing support |
| What happens if funding or a contract changes? | Communication, transition and access-ending arrangements |
| Can authorised records be retained or exported? | Clear scope, formats and conditions for the records the user needs |
| What happens when the service fails? | A support route and a practical fallback for the intended use |
| What evidence supports continued use? | Relevant evaluation, known limitations and how material changes are assessed |
The checklist is an original procurement and service-design aid, not a certification scheme. A supplier does not need to reveal every internal cost to answer the questions relevant to a user's continuity.
A fictional continuity exercise
Consider a training programme that encourages clinicians to use a free reference service and keep learning records. Before making that service central to its educational process, the programme could rehearse what happens if access changes.
Can learners still obtain their completed records? Are those records understandable without the original interface? Does the programme know which source was used for a particular learning activity? Can supervisors distinguish a learner's reflection from automatically generated text?
The exercise does not predict that the service will close. It identifies avoidable dependence. An organisation can value a free service while retaining a realistic route to continue its essential work if an entitlement, integration or funding arrangement changes.
The same reasoning applies to paid services. A subscription is not a continuity plan on its own.
How iatroX separates reference from paid learning
Per iatroX product information, September 2026, Ask-iatroX and free question access are genuinely free, without trial expiry or a verification gate. The paid subscription covers question banks, Socratic Tutor, the study planner, iatroX Simulations and CPD tools together. Simulations and CPD are included, not separately priced add-ons.
The published September 2026 price is £99 paid upfront annually, equivalent to £8.25 per month billed annually, or £29 per month. The upfront payment matters: £8.25 is not a monthly-payment option. At those prices, three monthly payments total £87 and four total £116, so the annual option becomes cheaper from the fourth month of paid use.
That does not make annual payment the right choice for everyone. Someone who only needs a short period of access may spend less monthly. Someone who only needs free reference may not need a subscription at all. The value case is several useful learning methods for one relevant professional goal, not a claim that access to unrelated examinations creates value for every reader.
Per the same September 2026 specification, iatroX CPD supports personally reviewed records, PDF export and direct FourteenFish export for linked accounts; completed evidence remains accessible after the subscription ends. Those are stated continuity features, not proof of long-term business sustainability or automatic accreditation of the learning.
Judge the arrangement, not the slogan
A free reference service may be an appropriate choice for an individual who needs accessible, source-linked information. An institution may reasonably pay for implementation, local content and support. A learner may pay for structured practice while continuing to use free reference.
These choices can coexist. The strongest arrangement is not necessarily the one with the lowest visible price, but the one whose entitlement, responsibilities and continuity match the intended use.
Sustainable access requires a service that can keep doing the work users rely on. Transparent pricing is part of that picture. Clear sources, maintenance, support and exit arrangements complete it.
Frequently asked questions
Does free medical AI mean there must be a hidden cost to the clinician?
Not necessarily: institutions, philanthropy, partnerships or paid adjacent services can fund genuine free access. The relevant questions concern the published entitlement, funding responsibilities and continuity, not an assumed hidden motive.
Are the financial arrangements of the Anthropic and OpenEvidence partnership known?
A complete financial account is not established by the public reporting reviewed for this article. The cost and funding models discussed here are a general framework, not a reconstruction of that deal.
When is iatroX's annual subscription cheaper than monthly payments?
At iatroX's published September 2026 prices, the £99 upfront annual subscription is cheaper from the fourth month, because four £29 payments total £116 while three total £87. The subscription includes banks, Tutor, planning, simulations and CPD, while free reference and free question access remain available separately.