If you've ever typed your symptoms into an AI chatbot at 11pm instead of waiting days for a GP appointment, you're not alone. But how much should you actually trust what comes back?
New polling from the Pew Research Center, surveying more than 5,000 US adults, suggests that most people still turn to healthcare providers when they want reliable health information - and that trust in AI for medical advice, while growing, remains firmly limited. The picture in Britain looks strikingly similar.
What the polls actually show
The data is striking for how clearly it reveals where people still place their trust. Around 85% of those surveyed said they get health advice from a provider, and 95% of those found that advice extremely or somewhat accurate. AI chatbots were used by only 22% of respondents - considerably fewer than the roughly 60% who turn to major medical organisation websites such as the Mayo Clinic or Cleveland Clinic, of whom 90% found them accurate.
That doesn't mean AI always gives poor answers. Among those who did use AI for health information, 76% said they found it accurate - a higher rate than social media, where only 51% of users said the same. The more telling finding is that most people aren't turning to AI at all - and those who are tend to be younger, uninsured, or without easy access to conventional care.
A parallel US poll found declining enthusiasm overall. Only 42% of the 1,007 adults polled said they were open to AI being used as part of their care, down from 52% when the same survey ran in 2024.
Why this matters in Britain
The NHS is under pressure, GP appointments are difficult to come by, and health information has never been more accessible online. In that environment, AI tools and symptom checkers have become a popular first port of call. But polling suggests British attitudes closely mirror the American data.
The Health Foundation's 2025 survey of 8,000 UK adults found that only 38% of the public believe AI will improve the quality of healthcare, with 19% saying they think it will make care worse. Notably, only 49% said they would be willing to use an AI-powered health assistant for non-urgent care, with nearly a third saying they would not.
The public's preference is also for oversight, not speed. 70% of Health Foundation respondents said they would rather AI outputs be checked by a human than returned faster without review, and 72% said they want stronger safeguards over potential benefits like convenience.
Where AI can genuinely help
That caution doesn't mean AI is useless for health - it's a question of what you're using it for. Among those in the poll who do use AI for health decisions, the most common use was understanding symptoms before deciding whether to seek medical care (62%), followed by making sense of test results or a diagnosis (44%), and preparing for an upcoming appointment (20%).
That pattern reflects where AI is probably most genuinely valuable: as a first pass, not a final answer. Helping you understand what a result might mean, framing a question ahead of a GP call, or working out whether a symptom warrants urgent attention are all reasonable uses. Arriving at a diagnosis is quite another matter - and the data suggests most people already sense that distinction.
Why trust in doctors hasn't gone away
The Pew poll found that 75% of Americans consider it highly important that their health information sources have medical training. That figure should dispel any notion that people are rushing to replace their GP with a chatbot. They aren't.
Ipsos polling of more than 3,700 GB adults in 2025 found that 52% prefer the NHS to continue using human-led triage systems, with many citing trust in human judgement and a desire for personal interaction. Women in particular favoured the human-led approach - 59% of women versus 44% of men preferred it. The research consistently points to the same underlying preference: people want AI working alongside clinical care, not instead of it.
What the research suggests you do
The polls don't say AI has no role in health - they say people are learning, sensibly, what that role is. Using AI to help you understand a test result, frame a question for your doctor, or work out whether a symptom is urgent: those are reasonable uses. Relying on it for a diagnosis or a treatment decision is a different matter entirely. Trusting that instinct, it turns out, puts you firmly in the majority.