Buckinghamshire Healthcare NHS Trust

Ryan Kerstein

Ryan Kerstein is the Associate Medical Director for Innovation and Research at Buckinghamshire Healthcare NHS Trust. He is also a Consultant Plastic Surgeon specialising in skin cancer. He is driven by a deep passion for improving healthcare through innovation, collaboration, and technology.

Ryan Kerstein’s career blends clinical excellence, entrepreneurial success, and strategic leadership, all centred on advancing healthcare innovation and shaping the future of medical practice. He shared his expert insights and thoughts for the 2025 edition of Healthcare Tech Outlook Europe.

Is AI Destroying Medical Research?

I was recently invited to deliver a keynote at an international surgical conference on artificial intelligence in surgical research. It was a timely opportunity, as the subject aligns with my roles: as a practising surgeon, Research Lead for my hospital, and editor for an international surgical journal.

Preparing that talk took me down a rabbit hole. Initially, I was convinced that AI would be an unqualified force for good in research, bringing efficiency, scale, and speed. Yet the deeper I looked, the more unsettling the picture became. By the end, I found myself ready to pose a more uncomfortable question: Is AI destroying medical research?

The Mirage of Progress

On the surface, medical research has never looked healthier. Around 1.5 million healthcare publications appear each year, and the WHO estimates biomedical articles double every 15 years. AI tools add fuel, generating abstracts, running analyses, and drafting full manuscripts. To the casual observer, this looks like extraordinary progress.

Yet volume is not value. The flood of publications risks drowning out high-quality science. Clinicians struggle to keep up. Peer review, the safety net of publishing, is stretched thin, with reviewers facing fatigue and time pressures. Generative AI adds another twist: it can create polished, plausible text with no guarantee of accuracy.

The result is a system that risks mistaking quantity for quality. AI can scale both insight and misinformation. Without safeguards, the scientific record could become saturated with content that looks credible but rests on shaky foundations.

AI as an Accelerator, not a Fix

There are genuine positives. AI can scan vast literature in minutes, flag patterns humans might miss, automate data extraction, identify trial cohorts, or assist in systematic reviews. Natural language processing is already used to analyse health records, spotting subtle signals for future trials.

Used wisely, these tools free researchers to focus on interpretation and judgment. For a junior researcher, AI can feel like a lifeline.

“AI can scale both insight and misinformation. Without safeguards, the scientific record could become saturated with content that looks credible but rests on shaky foundations.”

But AI is not a fix for deeper cracks. Poor design, publication bias, and incentives to publish quickly persist. AI may even accelerate these flaws. Garbage in still means garbage out, only faster and in greater volume.

The Risks of Synthetic Science

We are beginning to see “synthetic science”: articles generated by AI tools with minimal human oversight. Some slip into journals unnoticed, others flood preprint servers. There have been cases where AI-generated references were fabricated. For editors and peer reviewers, spotting these artefacts is becoming increasingly complex.

The risk is that the boundary between evidence and invention blurs. For clinicians and policy-makers who rely on research to make decisions, this is not trivial. If AI multiplies noise faster than we can separate the signal, patient care could suffer.

An even greater danger lies in the feedback loop: AI tools trained on the literature may reproduce AI-generated content. This “AI-cho chamber” risks compounding errors and amplifying misinformation. The more synthetic material enters the record, the greater the chance subsequent AI models will recycle it, creating a self-perpetuating cycle of distorted science.

Europe’s Opportunity

While these challenges are global, Europe is well placed to lead a more responsible path. Regulatory frameworks like GDPR already shape how data is handled. Initiatives like the European Health Data Space are creating secure, interoperable environments for sharing health data. Academic–industry collaborations are strong, and there is a growing recognition that AI requires governance as well as enthusiasm.

A coordinated European approach to “responsible AI in research” could provide checks and balances to harness benefits without succumbing to risks. Just as the European Medicines Agency sets standards for drug approval, there is scope to create similar frameworks for AI tools, covering transparency, data provenance, and reproducibility.

The Role of the Clinical Innovator

Central to this challenge is the workforce. AI will not replace researchers, but researchers who understand AI will replace those who do not. Clinicians must become intelligent customers of these tools, able to question outputs, understand limitations, and integrate insights responsibly.

This is why I believe in the role of the Clinical Innovator: professionals who bridge practice, research, and technology. At my hospital, through the BRAIn programme, and nationally via the Royal College of Surgeons' Innovation Hub, we are developing this capacity. Our aim is to ensure clinicians are not passive recipients of technology, but active shapers of its design and application.

Upskilling clinicians in AI literacy is essential if we want AI to serve research rather than distort it. This means embedding AI training in medical education, supporting fellowships in innovation, and creating time and space for clinicians to engage with technology.

Reimagining Research

For all the risks, there is immense opportunity. AI could help us move from static publications towards living reviews that update in real time. It could support personalised research, where insights are tailored to individual patients. It could accelerate discovery of patterns no human could identify unaided, unlocking the "unknown unknowns" with potential to transform healthcare.

In other words, the same AI that threatens to create an echo chamber can help break free from old paradigms—but only if approached with purpose, transparency, and humility.

Conclusion: A Call to Reflection

So, is AI destroying medical research? Perhaps not yet, but it tests foundations. It exposes fragility in systems that rely on quantity over quality, on peer review stretched to breaking point, and incentives rewarding speed over substance.

The choice is ours. AI can accelerate problems we face or help redefine research for the better. That will depend on equipping the workforce, creating proper standards, and insisting on responsible adoption.

AI will not end medical research. But unless we act thoughtfully, it could end research as we know it. With the right approach, it may help build a research ecosystem that is dynamic, personalised, and fit for the future of healthcare.