The Witness Gap
Sonya Cullington is a cyberpsychologist and digital policy advisor. She works on what sustained AI use does to human cognition, professional identity, and judgement, and what organisations and individuals can do about it. She is the creator of the AI Judgement Framework, the founding architect of the NHS Communications AI Taskforce, and chair of the Patient and Public Advocacy Steering Committee at UK Digital Health and Care.
For decades, health communications operated within a broadly stable dynamic. Institutions held the authoritative narrative. The job of communications was to shape, translate, and deliver that narrative in ways patients could act on. The craft was considerable, but the direction of information was largely consistent: from system to patient.
Generative AI has broken that model, and the sector has not yet registered what that means in practice.
Patients are no longer arriving at clinical encounters having searched for information. They are arriving after a personalised, validating, conversational process that feels nothing like a search. This is not accidental. The information environment patients now move through is designed to produce exactly this effect. Algorithmic content systems are optimised for engagement. In health contexts, engagement means personal resonance.
When a patient engages with generative AI or algorithmically curated content, they are not consuming health data. They are experiencing psychological mirroring. The algorithm does not just find a match for their symptoms, it also matches their narrative, creating a Self-Reference Effect: a conviction that the information is uniquely true for them because it reflects their lived experience so precisely. Information encoded in relation to the self is not just more easily remembered, it is more deeply believed.
Generative AI compounds this through the conversational interface itself. A patient who has spent time describing their symptoms to an AI tool has not just read an article, they have had a personal interaction. The tool has asked questions, reflected their experience back to them, and offered a framework that makes sense of what they have been feeling, in language that feels considered and specific to them. This creates Social Presence: the sense of engaging with a persona rather than a database. The psychological effect is closer to a trusted conversation than an internet search.
Platforms like TikTok add a further dimension. Patients are not just receiving personalised content. They are seeing themselves reflected in communities of people who describe the same experience, use the same language, and report the same journey from symptom to diagnosis to treatment so identity and belief become entangled. What can look like certainty in the consultation room often originates somewhere more complicated: in a patient who has felt unheard by an overstretched system, and who found, in an algorithm, something that finally seemed to listen.
By the time that patient sits in a clinical consultation, they may have already determined the diagnosis, identified the treatment they need, and prepared a response for when the clinician disagrees because what the patient is now defending is not just a piece of information they read or saw online, but a relationship they have formed with a system that made them feel understood.
The clinical encounter is no longer simply a moment of information exchange. It is a negotiation between two differently constructed certainties.
What many patients are seeking, before assessment, before diagnosis, before the pathway begins, is a moment of witness: the experience of being seen as a person rather than processed as a case. Communications teams bring exactly the relational and behavioural expertise that pathway design currently lacks. The challenge is that the environment they are now operating in has no established playbook. Understanding how patients form beliefs through algorithmically personalised content, how identity becomes entangled with health narratives online, and how conviction is produced rather than just information: none of this sits within existing communications training or professional frameworks. The craft is considerable. The context has changed faster than the curriculum.
The same is true for clinicians. Clinical training equipped them to hold expertise, navigate uncertainty, and build trust across a consultation. It did not prepare them for a patient who arrives having already negotiated their diagnosis with an AI system that made them feel genuinely heard. The clinical skills are there. The consultation dynamic has changed in ways that postgraduate education has not yet caught up with.
The patient arriving at the end of an AI-enabled pathway is carrying a belief system shaped by interactions the pathway designers did not account for. The designer assumes a clinical context. The patient is operating in a narrative and identity one. That is the context collapse at the heart of current implementation thinking, and it will not resolve itself.
When you are designing and delivering AI-enabled pathways, consider these questions, as they deserve direct answers.
Does your current pathway give patients a chance to feel heard before the clinical assessment begins?
When a patient arrives certain of their diagnoses, are you trying to correct their misinformation or understand what certainty means to them?
Are your digital pathways designed for a patient who wants to get through the system efficiently, or for one who is anxious and looking for reassurance?
To find out more, visit sonyacullingtonconsulting.com or connect with Sonya on LinkedIn.
