When AI Shapes Patient Expectations

Ms Anita Kemp and Dr Gaynor Radley discuss a case report from the Medical Protection files on how artificial intelligence is changing what patients expect

Artificial intelligence (AI) is increasingly influencing how patients understand health information and what they expect treatment can achieve. While AI tools can support patient engagement, they can also create unrealistic expectations that challenge the consent process and the doctor-patient relationship. This case highlights the importance of managing expectations, maintaining clinical judgement, and recognising when it may be unsafe to proceed.

Case study

Dr K, an experienced consultant cosmetic surgeon, was approached by Mrs F, who was seeking a blepharoplasty. At the first consultation, Mrs F explained that she was unhappy with the appearance of her upper eyelids and felt that they appeared heavy and droopy, affecting her confidence and mental wellbeing. She described the changes she hoped surgery would achieve.

Dr K listened carefully and assessed both her anatomy and her desired outcome. She explained what surgery might reasonably achieve, as well as its limitations. A digital scan was taken to produce a visual simulation, and the available surgical options were discussed. Dr K made clear that the simulation was illustrative rather than a guarantee of the final result. Mrs F appeared happy with the consent discussion, the details of which were carefully documented.

Two weeks later, Mrs F returned to review the simulation and became visibly upset. She insisted that it was markedly different from what had been discussed and asked the doctor to telephone her daughter. Dr K reviewed the scan with Mrs F and restated the proposed approach. Mrs F later returned with her daughter, apologised for her earlier reaction, and attributed it to anxiety. She reassured Dr K that she had confidence in her care and wished to continue.

Mrs F subsequently confirmed by email that she was satisfied with the proposed plan and wanted to proceed. Despite a sense of unease, Dr K continued to provide care. At another appointment, she and her daughter reviewed the simulation and again expressed their approval. Dr K reiterated the best outcome she believed could realistically be achieved.

Four days later, however, Dr K began receiving a series of emails. Mrs F had been comparing her appearance with photographs of strangers and celebrities online and sent screenshots highlighting individual features she wanted to replicate. She also supplied ‘before-and-after’ AI-generated images showing dramatic eyelid rejuvenation outcomes and recommendations from an AI chatbot, which suggested that the changes she wanted were entirely achievable.

At the next review, Dr K explained that the images were not a reliable representation of what could be achieved for her. Mrs F replied: “But AI said it was possible.”

That exchange crystallised Dr K’s concern. Although Mrs F had repeatedly expressed agreement, her desired outcome remained unstable and increasingly detached from the anatomical and clinical limitations that had been explained. She no longer felt that there was a sufficiently clear or durable shared understanding on which to proceed safely.

How Medical Protection helped

Dr K contacted Medical Protection. Having reviewed the chronology, the advice was to pause the proposed treatment and explain clearly that she was not prepared to proceed because the requested outcome was not clinically realistic and the divergence in expectations created a significant risk of dissatisfaction and harm.

Dr K was advised to offer Mrs F the opportunity to seek a second opinion and, if appropriate, to provide details of alternative specialists. She was also advised to record her clinical reasoning, the discussions about limitations, the changing requests, the AI-generated material, and the basis for ending the treatment relationship. Depending on the contractual arrangements and the work already undertaken, a refund of some or all fees could be considered as a goodwill gesture, rather than as an admission that the earlier care had been inappropriate.

The outcome

Mrs F declined the offer of a referral and preferred to identify another clinician herself. The parties agreed that her initial consultation fee would be refunded. The matter concluded without a formal complaint.

The professional position

The Medical Council’s Guide to Professional Conduct and Ethics for Registered Medical Practitioners states that the consent process requires dialogue, ongoing sharing of relevant information, and careful attention to the patient’s own wishes, understanding, and priorities. Patients should be given sufficient time to consider information before making a decision, particularly where the intervention is elective.

For HSE and HSE-funded services, the HSE National Consent Policy similarly describes consent as an ongoing process rather than a one-off event. It stresses that a signed form is only evidence that a communication process has taken place; it is not a substitute for a meaningful process. Although the policy does not govern private settings, it remains a useful benchmark. In this case, the repeated changes in Mrs F’s requested outcome meant that earlier expressions of consent could not simply be treated as permanent authorisation to proceed.

The Medical Council guidance also supports a doctor who decides not to provide a requested intervention where it is not clinically indicated or may cause more harm than benefit. It advises that the doctor should explain the reasons, offer referral for a second opinion, and document the grounds for the decision. Stopping an elective procedure in these circumstances is therefore not a failure of care; it may be the safest and most professionally defensible course.

Key learnings

► Explore the patient’s motivations, priorities, and understanding. Significant distress, repeated changes in the desired result, or reliance on idealised images should prompt a pause and further discussion.
► Describe digital simulation images as illustrative only. Avoid language that could be perceived as a promise or guarantee.
► Explain limitations in patient-specific terms, including the influence of anatomy, healing, scarring, and natural variation.
► Treat consent as a continuing conversation. Revisit it whenever the treatment plans, the patient’s expectations, or the information available changes.
► Patients have the right to have an advocate of their choice present during discussions about their healthcare and this can be helpful when undertaking a consent process.
► Do not allow AI outputs to displace clinical judgement. Explain that generative AI can produce plausible, but fabricated or anatomically unrealistic content.
► Where you decide not to proceed, communicate the decision respectfully, explain the clinical basis, offer a second opinion where appropriate, and make safe arrangements for transfer of relevant information with the patient’s consent.
► Keep contemporaneous records of the patient’s stated goals, the simulations shown, risks and limitations discussed, questions asked, emails and images received, and the rationale for any decision to decline treatment.
► Contact your medical defence organisation early – in this case, seeking support from Medical Protection enabled Dr K to navigate the situation before it escalated.

The Medical Council guidance also supports a doctor who decides not to provide a requested intervention where it is not clinically indicated or may cause more harm than benefit