AI-Generated Medical Advertising under NMC Guidelines 2026

Read the opening words of clause 7.2 first

The October 2026 guidelines use broad prohibitory language about AI-generated promotional campaigns advancing commercial interests. The subclauses address misleading AI representations, fabricated patient materials, disclosure when alterations may affect the audience, and privacy obligations for data entered into AI tools. A compliant-content marking mentioned within the clause creates an interpretative question; it should not be treated as eliminating the opening restriction. S1

This matters because two broad statements would both be unsound: 'every AI tool is banned in medicine' and 'all AI advertisements are allowed with a disclosure'.

Map the prohibited depictions

Clause 7.2(a) addresses misleading or unverifiable diagnosis, clinical outcomes, qualifications, patient experience and professional services. Subclause (b) addresses manipulated patient images, voices, testimonials, outcomes and synthetic endorsements intended to portray real individuals or experiences. Subclause (c) concerns appropriate disclosure of materially artificial content; (d) concerns the handling of patient information entered into AI systems. S1

The underlying concern is not a particular editing application. It is whether the communication deceives, commercialises professional practice in a restricted way or violates patient privacy.

Example: a synthetic IVF success story

An agency generates a convincing adult face and voice telling viewers that a named clinic guaranteed a successful pregnancy. No such patient exists. It fabricates experience and clinical outcome, and the content could also raise specific ART-sector legal issues. A watermark saying 'AI-generated' would not make the fictional result real.

Example: clinical anatomy animation

A clinician uses a scientifically verified animation to explain a coronary artery procedure without promising an outcome, portraying a real patient or directing viewers to book. This has a different factual and educational character from a synthetic success advertisement. Even then, verify its purpose, material accuracy, licensing, AI involvement and the scope of clause 7.2 before distribution. This example is not regulatory authorisation.

Example: AI editing of an actual interview

Noise removal, caption transcription and background correction do not necessarily manufacture a medical fact. But if voice cloning changes what a doctor said, visual manipulation suggests a result not actually achieved, or footage is used commercially, the risk profile changes. Do not treat 'minor edit' as a statutory exemption defined by the notice.

Data exposure through AI tools

Upload of imaging, clinical notes, patient photographs or identifiable metadata to an external AI provider raises confidentiality and processing questions independent of the final published output. Assess whether the tool retains inputs, trains on them, permits third-party access or sends them across borders. Apply present law and keep preparations for forthcoming DPDP duties distinct. S8 S9

A practical AI content register

Record the tool used, type of transformation, source asset and rights, whether real patients or clinicians are represented, whether claims are supported by medical references, any necessary disclosure, commercial purpose and human approvals. Restrict patient input data by default and escalate promotional campaigns to specialist review.

What creators can do now

Focus first on general, non-promotional, verifiable medical education; use original clinician explanations and independently licensed illustrations; avoid fabricated testimonials and treatment outcomes; and maintain an audit trail. The positive direction is to make healthcare information easier to understand without manufacturing clinical evidence.