By Jon Buckley, Head of Experience, Inizio Evoke Europe
In the age of AI, the answer itself is becoming a central part of the healthcare experience
At Inizio Evoke Europe, patients are at the heart of our mission to make Health More Human™. That’s why the focus of World Patient Safety Day on September 17, “Safe care for life,” for people living with noncommunicable diseases (NCDs), resonates so strongly with our team.
74% of deaths worldwide are caused by NCDs, with the World Health Organization estimating that 1 in 10 patients experience harm during healthcare. For some conditions, including cancer, adverse events can happen in as many as a third of patients. A sobering thought, but it’s not inevitable – around half of this harm is preventable.
One of the WHO's most interesting messages this year is about going beyond an individual treatment or intervention to ensure this patient safety. People living with long-term conditions move between diagnosis, treatment, monitoring, self-care, different professionals and different healthcare settings. Safety must travel with them, into every aspect of their daily lives.
So, how can this be implemented, and in this day and age, what part can AI play?
AI is already firmly embedded in the healthcare experience
AI is no longer on the horizon or being spoken about in hypothetical terms. The future is well and truly here. A US survey conducted in May 2026 found that 29% of adults said they were using AI tools or chatbots for health information at least once a month. Two years earlier that figure was 17%, so things are changing fast.
Given these figures, it’s unsurprising that between appointments, people frequently ask AI about symptoms, diagnoses, medicines, side effects and what their next steps should be. And why wouldn’t they? Whether it’s translating something they didn't understand in the consulting room, preparing questions for their next appointment, interpreting a test result, or simply because it’s 11 o'clock at night and their doctor isn't available, AI can step in. Whether or not healthcare organizations intended for this to be part of the current patient experience, it very much is. But this creates a very interesting paradox.
AI can sometimes feel more human than the system around it
AI is often presented as a threat to the humanity of healthcare, but the experience seems to be more complicated than that.
In a 2023 JAMA Internal Medicine study, healthcare professionals evaluated physician responses versus ChatGPT responses in answer to 195 patient questions posted on Reddit. The chatbot responses were preferred in 78.6% of evaluations and were rated significantly higher for empathy.
Admittedly, the study had important limitations. These weren't consultations. Patients themselves weren't doing the judging. And the AI answers tended to be considerably longer.
But I still think the finding is intriguing. It doesn't tell us that AI has empathy, but it does tell us something about what empathetic communication can look like: enough time, clear language, acknowledgment of the question, and a response that doesn't feel rushed.
In an ever-stretched, time-poor healthcare system, AI may be able to provide more of those things. But there is an important catch.
Be wary of unchecked confidence
Like a larger-than-life character in a local pub, generative AI is extraordinarily good at sounding as if it knows what it is talking about. Unfortunately, sounding right and being right are two different things.
The WHO has warned that large multimodal AI models can produce false, inaccurate, biased or incomplete health information. And because an answer can still sound confident and plausible, errors may be difficult for users to recognize. ‘Automation bias’ is also a watch-out, where patients or healthcare professionals trust an automated output enough to overlook something they might otherwise have questioned.
It’s no secret that all of this really matters when the subject is health. In research conducted by KFF this year, 44% of adults said they weren’t confident gauging whether health information from AI chatbots was true or false.
There are subtler experience problems too. An answer can be clinically plausible but wrong for the healthcare system someone lives in. A treatment pathway that makes sense in the US may make little sense to someone navigating the NHS. Language can be accurate but inaccessible. And information can be technically correct without giving someone enough context to know what they should do next.
Another study, by JAMA Network Open, found that AI-drafted patient messages were perceived as being more empathetic, but were also more linguistically complex and less readable than human-written messages. So, a good AI answer and a good healthcare experience are not necessarily the same thing.
Patient safety is a key experience design challenge
When I say patient safety is our challenge, I don't mean experience designers should start making clinical decisions. I mean that the decisions we make about an experience can affect what somebody sees, understands, trusts and ultimately does next.
And that brings us back to patient safety. If AI is part of the health experience you’re creating, then trust becomes a design responsibility. When patients’ lives are at stake, that trust has to be earned through accuracy, transparency and a clear understanding of where AI ends and human judgment begins.
We need to start by asking five important questions.
1. How do we make sure the answer stays accurate?
Approved healthcare content has already been through medical, legal and regulatory review. The new risk is what happens after that point: AI can summarize it, combine it with other sources or apply it outside its intended context. We need to design for accuracy, traceability and clear boundaries around what AI should and shouldn’t answer.
2. Is it right for this person, in this place?
Clinical accuracy isn't enough. Healthcare system, language, culture, accessibility, health literacy and individual context all matter.
3. Does the experience make any uncertainty obvious?
People should be able to understand what is known, what isn't, and where information has come from.
4. Where does AI stop and human judgment come back in?
There need to be clear boundaries. Personal medical questions, treatment decisions, concerning symptoms and situations involving risk should not simply generate another AI answer. The experience needs to recognize when to stop and signpost the person to the appropriate healthcare professional or service. If a company-controlled interaction raises a potential adverse event or product-safety issue, it also needs to be routed into the appropriate pharmacovigilance process, rather than simply continuing an open-ended conversation.
5. Have the people whose safety depends on it helped design it?
This last point feels particularly important this World Patient Safety Day. One of the WHO's explicit 2026 objectives is to involve people living with NCDs in identifying safety risks and co-developing solutions alongside healthcare professionals, leaders and policymakers.
For anyone designing patient or HCP experiences, that should be a non-negotiable design principle, not a ‘nice to have’.
And there’s another implication for healthcare brands. Most pharmaceutical companies aren’t going to be building the general-purpose AI systems their audiences use. But they will be creating some of the information those systems discover, interpret, summarize, compare and surface. This means looking beyond AI product design, to information design too.
For me, this is where Generative Engine Optimization and Answer Engine Optimization become far more interesting tools for healthcare. GEO/AEO can’t simply be a visibility exercise. It should be about making appropriately approved information clear, authoritative, contextual and discoverable, giving AI systems the best possible source information from which to construct an answer. And where a healthcare company controls the experience, the design also has to respect regulatory boundaries around promotion, personal medical advice and product safety. Because if the answer is already part of the healthcare experience, the ambition shouldn’t simply be to be part of it. That’s a given. It should be to enrich the whole experience to benefit the patient. AI has enormous potential to make long-term care easier to understand, more accessible and, perhaps surprisingly, more human.
But when AI can give us answers, humans still have to engage critical thinking to own what happens next. And when it comes to focusing on ‘Safe care for life’, it feels like real food for thought for all of us designing healthcare experiences.
Interested in hearing more? Connect with us here.