AI in healthcare: adopted, trusted – and still unproven

One in seven people in the UK have used an AI chatbot for health advice instead of seeing a GP.It’s a striking figure – but the reason behind it is just as telling. King’s College London, reporting new UK survey findings in May 2026, found that 25% of those who had turned to AI for health advice did so because they had waited too long for NHS services.

That changes the nature of the conversation. The question is no longer whether AI will be part of how people access health support. It already is – in public behaviour, in procurement decisions, and in the government’s 10 Year Health Plan, which puts the shift from analogue to digital at the heart of NHS reform. The real question is whether we can say with confidence that it’s working.

And adoption isn’t the same as evidence. A tool can be widely used and still fall short on safety, fairness or effectiveness. The same King’s study found that one in five people who used AI for health advice said it discouraged them from seeking professional care. That deserves serious attention.

The governance gap

The NHS is committing substantial resources to AI. A new £900m framework for healthcare AI solutions was launched this month. But governance is still developing. A December 2025 pollfound that approvals and evaluation readiness are now the bottlenecks across NHS organisations responsible for AI procurement. The decision to buy can move faster than the capacity to assess.

We’re making progress: the MHRA is due to publish a new regulatory framework for AI in healthcare later this year, informed by the National Commission into the Regulation of AI in Healthcare.

In the meantime, though, the speed at which AI tools are coming to market means governance frameworks are struggling to keep pace – leaving NHS organisations to navigate inconsistent standards and duplicative internal processes.

Is AI improving health equity?

AI could also create potential health inequalities. People are turning to AI partly because they cannot get timely access to NHS services. If AI fills a gap created by lack of system capacity, it risks becoming a two-tier arrangement – useful for those confident enough to navigate it, and out of reach for those who are not.

The evidence is growing. A May 2026 paper in the New England Journal of Medicinewarned that a digital divide affecting financially, racially and geographically marginalised groups may widen as AI becomes more embedded in health care.

Research published in the European Heart Journal found that when AI algorithms are deployed at scale without validation across diverse populations, existing disparities risk being amplified. And NHS England’s own datashows that uptake of digital health tools remains significantly lower in the most deprived communities, with 2.1 million people in the UK still entirely offline.

A tool being free and widely available doesn’t always mean it’s equitable. The question is who AI works for – and who it doesn’t.

Where evaluation fits

These are the areas where evaluation matters most. It’s the discipline that asks the questions procurement may not. Who benefits? Who is excluded? What happens to risk? Does the tool improve decisions, access or outcomes – or does it create the appearance of progress without the substance?

Those questions are especially important now. The NHS is under pressure to transform at pace. But as The King’s Fund has noted, ambition on paper and change in practice are two different things.

At Future Care Capital, evaluation isn’t something we apply after the fact. It is part of how innovation earns the trust it needs to scale. Because in health and care, widespread use is not the same as proven benefit.

If you’d like to discuss how Future Care Capital can help your organisation evaluate AI in a way that’s proportionate, credible and useful, drop an email to our evaluation lead, Professor Andy Jones, at andy@futurecarecapital.org.uk