Can you trust AI for health symptoms? Here’s what we found

You wake up with a pain in your side. But before you call your GP, tell your partner, or finish your tea, you open ChatGPT. Sound familiar? You’re not alone.

A growing number of us are now turning to AI chatbots to uncover the underlying cause of their symptoms. But while AI is quickly becoming part of the nation’s health routine, there’s one big problem:

Most people don’t actually trust what it tells them.

And even more worrying? Most aren’t confident they could tell when AI is wrong.

How many of us are actually doing this?

According to a new poll by Bluecrest, nearly one in five Brits (19%) use AI chatbots to check their health symptoms.

That might sound surprisingly high, but it still trails behind Dr Google. Almost half of adults (47%) still turn to search engines when something feels off, proving that “symptoms + NHS” remains one of the nation’s most-used search combinations.

But AI is creeping into healthcare habits in new ways too. Around 15% of people now use chatbots to prepare for a GP appointment, compared to just 11% who read medical journals or research papers beforehand.

The trust problem

Interestingly, despite more people using AI for health advice, hardly anyone fully believes it.

Just 12% of adults say they trust AI-generated health information.

Compare that with:

• 78% who trust the NHS website
• 71% who trust their GP

And even traditional Google searches outperform AI on trust.

AI tools now sit among the least trusted sources of health information in the UK, and only one in three people (33%) say they feel confident spotting incorrect information from AI chatbots.

That means millions of people are using tools they don’t entirely trust and aren’t fully equipped to fact-check.

Even transparency seems to make things worse. More than half of adults (56%) said they would trust health information less if they knew it had been generated by AI.

Meanwhile, over a quarter (27%) of AI users admit they only verify the information they get sometimes, rarely, or never.

That’s a risky mix: low trust, low verification, high usage.

What AI can’t do

AI is largely great at gathering information. It can explain symptoms, summarise conditions, translate medical jargon, and suggest possibilities faster than a search engine could.

But there’s something it fundamentally lacks: context.

It might know millions of pages worth of information about blood pressure, fatigue, or chest pain – but it knows nothing about your blood pressure, your fatigue, or your history.

AI doesn’t know:

• What your normal resting heart rate looks like
• Whether your cholesterol has changed over time
• If “always tired” is new for you or completely normal
• Whether this symptom has appeared gradually or overnight

And that matters. AI can offer general information, but it can’t replace personalised medical understanding.

The real issue might be our Health IQ

The findings also revealed something bigger than AI itself; many of us struggle with basic health literacy.

Separate analysis using Bluecrest’s Health IQ tool found the average British adult scores just 63.8 out of 110 on everyday health knowledge.

That technically places most people in the “average” category, but only just.

Nearly one in four adults (24%) still fall into the “developing” bracket, suggesting many people lack confidence in understanding common health information and risks.

Only 21% reached “expert” level knowledge.

And surprisingly, younger adults scored worse than people in their late twenties and early thirties. Those aged 18–24 scored an average of 57.5, compared to 68.6 in the 25–34 bracket. That matters because AI tools are only as useful as the knowledge behind the prompts.

If you don’t understand what you’re reading, it becomes much harder to spot misinformation and know what questions to ask. In other words, AI doesn’t eliminate health confusion; it can sometimes amplify it.

A postcode lottery for AI trust?

Where you live also appears to shape how likely you are to trust AI with your symptoms.

London stands out as the UK’s AI health hotspot, with roughly double the national level of chatbot usage. Londoners also reported greater confidence in spotting misinformation and higher overall trust in AI-generated advice.

Meanwhile:

• Northern Ireland showed unexpectedly high adoption and above-average trust
• Yorkshire and the Humber appeared more traditional, with stronger trust in GPs and lower AI engagement
• Wales recorded some of the lowest digital health engagement overall
• Scotland appeared more cautious around sharing personal health data
• The South East and South West reported the strongest privacy concerns

What actually helps?

The irony is that most people aren’t really looking for medical advice online – they’re looking for reassurance.

They want to know:

• “Is this normal?”
• “Should I worry?”
• “Do I need to see someone?”

And often, panic spirals because people have no baseline for what’s normal for them. That’s where understanding your own health data becomes genuinely useful.

Knowing your cholesterol, blood pressure, heart health markers, or vitamin levels gives context that neither Google nor AI can provide on their own. It turns random symptom-checking into informed decision-making. It also helps cut through the noise.

Because when you understand your own trends over time, you’re less likely to catastrophise every symptom or blindly trust the first answer a chatbot gives you.

So, can you trust AI for health symptoms?

Sometimes. Carefully. AI can be useful for explaining terminology or preparing questions for your GP, but it works best as a support tool, not a substitute for medical expertise or personal health awareness.

Because ultimately, the smartest health tool isn’t necessarily the one with the biggest dataset. It’s the one that helps you understand your own body better.

Take our Health IQ test today and find out how your score compared with the national average.