Blog: What becoming a patient taught me about AI

In January this year my appendix ruptured while I was on a skiing holiday in France. What followed was several months of complications, multiple hospital admissions, weeks of antibiotics and a severe reaction to medication that caused my immune system to collapse and damaged my liver.

As someone who works in digital health comms and who feels cautiously aware of the risks and unintended consequences of AI, I never imagined I would end up reflecting on the role it played in helping me navigate the most serious illness of my life.

‘AI wasn’t diagnosing me, it was influencing my behaviour’

After dismissing stomach pain for two days as indigestion from too much cheese, a 2am conversation with ChatGPT about French wind relief tablets suggested I should seek immediate medical help. This was the first point I realised AI wasn’t diagnosing me, it was influencing my behaviour. Within 24 hours and following the most excruciating pain of my life, I had undergone emergency surgery for a badly ruptured appendix.

Two weeks later and back home in Cardiff, I was still struggling. Persistent stomach pain and queasiness had me researching recovery timelines and possible complications. Google's AI summary suggested an intra-abdominal abscess as a possible explanation.

A late-night call to NHS 111 and a visit to A&E the following day resulted in reassurance and a diagnosis of normal post-operative recovery. I went home relieved.

But something didn't feel right and within days I developed fevers and worsening symptoms and returned to hospital, where scans revealed a sizeable abscess on my colon that had developed following my surgery. I spent the next week very unwell in hospital receiving urgent treatment and was eventually discharged on a prolonged course of potent antibiotics.

‘I interrogated the medical advice I received’

By this point I had become the sort of patient I never imagined I would be. I monitored every new symptom and interrogated the medical advice I received through my own research. I had AI tools available on my phone at all times, day or night, to help me make sense of what I was experiencing.

Several weeks later I developed another fever, a rash and a sore throat. My GP said I had likely picked up a virus. AI suggested something more concerning: a potentially life-threatening reaction to one of my antibiotics that required urgent medical attention.

Blood tests a day later showed that my immune system had almost completely collapsed. My neutrophil count had fallen to 0.1, leaving me highly vulnerable to sepsis, while my liver enzymes continued to rise dramatically. I was admitted back into hospital for more antibiotic treatment and close monitoring. I sent a WhatsApp from my hospital bed to my manager: “AI got it right again.”

‘AI highlighted possibilities that prompted me to seek help’

Fortunately, I was able to make a full recovery thanks to the amazing care I received from healthcare staff at the Heath, although it took another month for my liver function to return towards normal.

Looking back, I don't know whether AI saved my life but I do know it changed the decisions I made. Three times in three months, AI highlighted possibilities that prompted me to seek help at critical moments, ask questions or challenge reassurance that didn't quite fit what I was experiencing.

At the NHS Wales Digital Health Conference in May, several weeks into my recovery, I attended Gartner's AI breakfast session and one slide in particular stuck with me. It described the evolution of the much-discussed phenomenon of ‘Dr Google’ into ‘Dr Google AI Summary’. The implication was clear - patients are increasingly turning to AI-generated health information, and that should make us stop and think.

I found myself staring at the slide thinking, that's me. Not as a future trend or a hypothetical risk, but as a patient right now.  

For months I had been using ChatGPT and Google's AI summaries while navigating complicated and serious health issues. On more than one occasion, the information I received prompted me to seek medical attention or take symptoms more seriously than I otherwise might have done, and that leaves me feeling conflicted.

‘It also helped me to interpret information and understand terminology’

AI can be wrong, it can hallucinate, it can misunderstand context and it can provide poor advice with remarkable confidence. There are also wider questions about transparency, accountability, privacy, security and environmental impact that we are only just beginning to grapple with. 

What struck me most was not that AI was acting like an on-hand clinician, it wasn't. What it provided was something different – it helped me interpret information, understand clinical terminology and to decide whether a symptom was something to monitor, something to discuss or something to act upon. It helped me make sense of a healthcare journey that spanned multiple countries, languages, organisations, services and clinical teams.

It also felt like the only constant I had – a steady, immediate source of information in the darkest and scariest moments. That particular role for AI, helping people interpret and navigate care, feels increasingly significant.

‘But AI could only help me because information existed in the first place’

As both a patient and someone who works in digital health, it raises some important questions for me. If these tools genuinely help people navigate complex healthcare journeys and advocate for themselves more effectively, what does that mean for digital inclusion? What happens to people who don't have access to them, don't know how to use them or don't trust them? How do we help people understand when AI-generated information is useful and when it isn't? 

Are we creating a new form of health inequality based not just on access to care, but on access to interpretation?

I am conflicted and uneasy about those questions, but I do know that AI could only help me because the information existed in the first place. Blood tests, scans, prescribing records and clinical correspondence all formed part of the picture. The less visible work of building and connecting digital services that we do at DHCW matters so much because it creates the foundation on which everything else depends.

This experience reminded me that the digital health journey we enable is far from abstract, it is about real people, often at moments of uncertainty and vulnerability, desperate to understand what is happening to them, what it means, and what they should do next.