ChatGPT missing gynaecological cancers when asked about ‘red-flag’ symptoms

Our testing of ChatGPT found serious concerns, with it missing gynaecological cancers as potential causes of leading symptoms, giving inaccurate or outdated information and giving inappropriate medical advice.

Misinformation and disinformation are a growing problem, with easier access than ever to information through the internet and social media, there is a growing number of unqualified people sharing potentially inaccurate health information online, more scams targeted people’s anxieties and worries about their health, and businesses trying to sell unnecessary or even harmful products. We wanted to see how AI platforms like ChatGPT did at filtering through the noise, and how much it could be trusted to give information on gynaecological health.

We ran 25 test scenarios with ChatGPT, where we posed as patients with key symptoms of the five gynae cancers.

What we found

In the test scenarios we ran, ChatGPT did not pick up ovarian cancer as a possible cause of persistent bloating for more than three weeks- one of the leading symptoms. While it did recommend seeing a doctor, it kept offering diet advice and tips around dealing with irritable bowel syndrome (IBS). IBS is a benign condition that patients with ovarian cancer are commonly misdiagnosed with. Ovarian cancer still wasn’t mentioned even when we explained that the hypothetical patient had already changed their diet and their bloating was getting worse.

Vulval cancer was also missed as a potential cause of vulval itching. Even when we gave details of vulval itching which exactly matched how it would present if it was a sign of cancer, like it hadn’t improved with over-the-counter medicines and was worse at night.

We also found that the information we were given by ChatGPT was often not relevant to the UK’s health system, or not in line with NHS guidance. ChatGPT sometimes provided outdated or inaccurate information, like that you are invited to cervical screening every three years in England from the ages of 25-49. This was changed to every five years in 2025. It also said that cervical screening was used to identify VAIN (vaginal intraepithelial neoplasia – cell changes in the vagina), but these cannot be detected by cervical screening.

In other scenarios, ChatGPT inappropriately gave advice that should only be issued by a registered medical professional, such as recommending a FODMAP diet for persistent bloating without asking whether the ‘patient’ had any other diseases that would make the diet unsuitable.

Why we are worried

We are particularly worried because more and more people are turning to ChatGPT and other platforms like it for health information and advice. We surveyed over 2000 women in the UK (YouGov, March 2026) and found that AI platforms have overtaken social media as places women go to with their gynae health questions and concerns.

NEARLY
1 in 10

(9%)

women who had experienced gynaecological health issues had sought information and advice from AI platforms like ChatGPT

1 in 4

(24%)

of these women said they ‘felt reassured that they didn’t need to worry’ when asking about their gynae health symptoms

LESS THAN
1 in 3

(28%)

said it left them feeling encouraged to see a doctor

ONLY
16%

said they didn’t fully trust the information they got.

Nearly 1 in 10 women (9%) who had experienced gynaecological health issues had sought information and advice from AI platforms like ChatGPT. 8% had gone to social media for information.

When using ChatGPT only around 1 in 4 (27%) had told it they were based in the UK and less than 1 in 4 (23%) of them had clicked on the source links to check the reputability and accuracy of the information.

“LLMs like ChatGPT and Claude rely on the information available online. With rare cancers, like ovarian and vulval, they can easily miss highlighting symptoms as a possible sign of cancer. ChatGPT wants to please the person using it, and it can give false reassurance. And if you’re on the receiving end, you might ignore the symptoms you were having instead of telling your GP. We need to remind people that AI cannot accurately diagnose you and it does make mistakes.”
-Dr Liz O’Riordan

Growing evidence

In the UK, the number of ChatGPT users more than doubled in a year – rising from 6.5 million in September 2024 to 16 million by September 2025. 66% of users say they turn to it for questions about their health and wellbeing.

Over the last few months, several research papers have been published which shine a light on similar concerns. One paper was published in Nature Medicine this February, which highlighted similar safety concerns around ChatGPT Health missing potentially life-threatening health emergencies in over half of the scenarios they posed.

The Patient Information Forum with IPSOS have also found that 88% of consumers in the UK are using online resources to gather health related information. 41% of us have been exposed to inaccurate or misleading health information online in the last three months. IPSOS found that patients are arriving at consultations pre-researched, guided by TikTok, ChatGPT, and health influencers.

Tips to spot misinformation

This May, to help you find trusted health information online, we are sharing a toolkit created by the Patient Information Forum:

Artificial intelligence (AI) is a term to describe computers that learn and carry out “human” tasks. AI has some good uses in health information but public AI tools come with risks.

1) How can AI be used in health information?

Here are some ways AI is used in health information:

Translating websites or apps into other languages
Powering chatbots on websites
Studying research data.

Good health information is open about using AI. For example it will say: “This website uses AI translation.”

2) What is generative AI (GAI)?

Generative AI (GAI) describes online tools or apps that give or “generate” answers to questions asked. GAI can create text, images and video. When asked a question or a “prompt”, a GAI tool will search data to give an answer. ChatGPT, Co-pilot, Gemini and BLOOM are examples of GAI.

3) Can I trust GAI to give me accurate answers?

Not at the moment. So be very careful if you use GAI to find health information.

4) What is the problem?

GAI might give answers that are out of date or share misinformation from the data it has searched. GAI can reflect racial, gender and other prejudices. If GAI cannot find useful data it can sometimes “make up” answers that sound real.

5) How can I use GAI?

Ask good questions:

Keep questions simple
Be specific
Ask GAI tools to search trusted information sources.

For example, “Find information on cervical screening in England. Search the NHS and health organisations with the PIF TICK.”

Fact check the answer given

Ask GAI where it got the information for its answer. But be aware it might not give this information. If it does, check the original website or other source.

Always check health advice with a health professional. You can speak to our Ask Eve Nurses for free and confidential information and support on gynae health. Do not act on advice from GAI tools.