Why do we continue to assume that patients themselves aren’t the best source for understanding the lived experience? Historically, healthcare professionals (HCPs) have often been used as proxies for patients, based on the belief that clinical expertise naturally provides insight into patient needs. However, research has consistently shown that clinicians tend to misjudge patient preferences, particularly when it comes to quality of life, treatment burden and adherence. This reflects a fundamental difference in perspective: clinical versus lived experience. As a result, decisions are too often based on incomplete, and at times inaccurate information. Today, we risk repeating the same mistake with artificial intelligence. While AI can aggregate and synthesise vast amounts of data, it ultimately reflects the information it has been trained on. This means existing gaps can be reinforced, potentially leading us in the wrong direction. Much like clinicians, AI can “know” about patients without truly understanding what it means to be one. Take Josie’s diagnosis journey, shared previously by our patient research team, Patient Panorama. Much of what she experienced was invisible to her HCP. It would be misguided to assume their perspective accurately reflected her reality. An HCP-led proxy approach would miss the emotional burden, frustration and sense of invisibility that shaped her journey. Without this understanding, there is a real risk of developing and funding interventions that fail to address the true challenges patients face, leaving many feeling overlooked or misunderstood. Ultimately, this means resources are allocated based on assumed patient profiles that don’t align with reality. AI may appear to offer an improvement. It can sometimes identify emotional themes such as dismissal, stigma and psychological impact. But sounding like a patient is not the same as being one. We must be careful not to become complacent. It’s important to recognise where AI falls short, and how those gaps can lead to interventions that overlook, or even misunderstand, entire groups of people. Although AI can add value by highlighting certain emotional aspects, it still doesn’t go far enough. We are not yet doing everything we can to genuinely improve patient experiences, whether that’s speeding up diagnosis, improving access to treatment, or ultimately supporting a better quality of life. The Limitations of AI as a Patient Proxy Loss of context and nuance Qualitative insight isn’t just about what is said, but how, when and why it’s said. AI can identify themes, but it often flattens contradiction, hesitation and emotional weight, the very elements that matter most in experiences such as not being believed. No true lived experience AI does not experience uncertainty, fear or power dynamics. It can describe dismissal, but it cannot feel how that shapes behaviour over time, for example, why a patient may delay returning to seek help. Gaps in representation AI is only as strong as the data it is trained on. That data often over-represents patients who are articulate, engaged and willing to share their experiences. Those who disengage, are silenced, or never reach diagnosis are far less visible, meaning existing gaps are not only preserved, but potentially amplified. Bias within AI systems We still don’t fully understand how AI arrives at its conclusions. While we know it depends on its training data, we don’t always see what patterns it is prioritising. A well-known example is Amazon’s abandoned AI recruitment tool, which showed bias against women by favouring candidates associated with male-dominated activities. In healthcare, a similar issue could arise if AI identifies who is most visible in the system, rather than who is most at risk. The model may appear accurate, but it could simply be reflecting patterns of access and documentation, not genuine clinical need. Moving Forward To avoid repeating past mistakes, direct engagement with patients must remain central. AI should be viewed as a tool to support understanding, not replace it. Combining AI-driven insights with in-depth qualitative research, patient-reported outcomes and behavioural data can help to address blind spots. There is, however, an additional challenge. We are all exposed to vast amounts of information every day, and naturally rely on mental shortcuts to process it. Much of our decision-making happens subconsciously, meaning people are not always able to articulate the factors influencing their behaviour. This creates further gaps that neither AI nor traditional methods can fully capture without careful exploration. If AI can only learn from what is visible and recorded, how do we design solutions for what is unseen, undocumented or unspoken, particularly in conditions shaped by stigma, such as endometriosis? The conclusion is straightforward: no matter how advanced, proxies cannot replace authentic patient voices. At the same time, effective market research requires more than simply speaking to patients. It demands expertise in uncovering what isn’t immediately visible, the gaps, the silences, and the unspoken realities of human experience. This is where the real value lies: helping organisations understand not just what patients say, but what they truly go through. Because ultimately, it’s not about abstract data or simulated insight. It’s about real lives. And that’s what allows decisions to be grounded in how people actually live and feel, not just in patterns of data To find out more the importance of the real patient voice in healthcare research, fill in the Contact form below Apply Now!