When AI answers your cancer questions: what you should know
Can you trust AI for cancer information? What patients need to know.
Searching for answers about your cancer online? Here’s what AI chatbots can, and can’t, reliably tell you, and how to use them safely.
If you're living with cancer, the silence between medical appointments can be the hardest part: you probably have lots of questions. Maybe you're feeling scared, overwhelmed, or just want to understand more about what's happening. This is completely natural. Today, many people turn to AI chatbots like ChatGPT, Copilot, Gemini, Claude, DeepSeek, Perplexity or Meta AI to find support, because they're available anytime, and they can seem like a quick way to get information.
In my work as a doctor at Anticancer Fund, I regularly speak with patients who come to me with questions, often ones they’ve already tried to answer online. Sometimes the information is helpful, sometimes it is worryingly wrong.
So, here's something important to consider: while these tools offer speed, the do not offer safety. Before you rely on AI-generated information, ask yourself: is this accurate? In the context of cancer, and cancer treatment, acting on incorrect data isn’t just confusing, it can even be harmful.
Beyond the algorithm: how AI ’thinks’ and why it differs from your medical team
To understand the risk, we must look at how these tools ‘think’. When you ask AI a question, it breaks your words apart. It matches them to patterns learned from billions of texts. Then, it builds an answer based on what it has seen before, on statistical likelihood, not medical training, or understanding of your unique situation.
Your healthcare team operates differently. They combine years of education, experience, and clinical judgment. While AI sees a pattern, they consider your biology, your symptoms, your test results, and who you are as a person. They know what matters most in your specific case, whereas AI is simply guessing.
Where AI can help: organising information and simplifying medical jargon
This doesn’t mean AI is without value; it simply requires the right context. AI chatbots can provide extensive and detailed information. They’re good at sharing general facts that don’t need personalisation or focus on a specific cancer type. For example, research shows that AI can give useful information about cancer prevention and survivorship. It tends to provide consistent answers to general questions.
One study on endometrial cancer showed AI provided complete answers 91% of the time, compared to 49% for doctors. This doesn’t mean AI is better than your doctor. It means AI may give longer, more detailed answers to general questions. However, in medicine, length doesn’t equal accuracy.
AI can also help you prepare for medical conversations. It can be an excellent tool to:
- Explain a complex medical term in simpler language.
- Help you list important questions for your next appointment.
- Organise your thoughts before meeting your care team.
AI is great at organising and simplifying complex information. I tried this by using AI to restructure this blog post for better readability. This highlights a key strength in the medical field: AI helps researchers and doctors cut through jargon. It ensures that important health information is clear, engaging, and easy for patients to act on.
Where AI falls short: hallucinations, bad advice and privacy concerns
AI has real strengths, but also real limits, especially when it comes to personal health. This is where I want to be direct with you, because in my clinical experience, this is where things can go wrong.
Several studies show that AI is less reliable for questions about cancer diagnosis and treatment decisions. For rare cancers, such as salivary gland cancer, AI provided incorrect answers in about 13% to 28% of complex treatment questions. This varied based on the model used, as shown in recent research. Even for simpler topics, like lifestyle advice for cancer survivors, AI’s performance can be poor. In one breast cancer study, fewer than 10% of responses about exercise recommendations were correct.
Here's what an AI-tool cannot do:
- It doesn’t know you personally
- It cannot examine you or review your test results considering your complete health history.
- It cannot take responsibility for medical decisions and it is not accountable for what it suggests.
- It cannot assess your individual risks and weigh the individual pros and cons of a treatment for you.
Three specific concerns worth knowing about
1. Confident wrong answers
AI sometimes makes mistakes and presents them as facts, a phenomenon called "hallucinations”. In one study about prostate cancer, when someone asked, "Which mushroom fights cancer best?", AI confidently replied, "The Turkey Tail mushroom is known to help with cancer.". This had no reliable evidence and included false information.
What to do: If AI gives you a specific claim about treatments or supplements, verify it with your care team before acting on it.
2. Answers that feels overwhelming
AI responses often use complex words and long sentences. They may seem clear at first but often lack practical advice. Research shows AI responses are usually longer than those from physicians, which can feel overwhelming when you’re already dealing with a lot.
What to do: If something feels confusing, pause and ask your nurse or doctor to explain it in plain language.
3. Your privacy may not be protected
Your private conversations with AI chatbots might not be secure. Studies show many AI companies save your messages, often without asking for your permission. This data could be stored, leaked, or shared accidentally.
What to do: Assume anything you type could be seen by others or used to train the AI unless the company states otherwise. Check the privacy settings and privacy policy before sharing personal information with any AI tool. Also, avoid entering your full name, date of birth, address, hospital number, or detailed medical reports into AI chatbots.
Practical steps to stay safe
It’s completely understandable to want quick answers when facing cancer. AI can be a starting point for learning, but it should never replace your healthcare team. They know you, are accountable for your care, and can provide guidance tailored to your situation. If you have questions or concerns, talk to them at your next appointment.
You can also reach out to My Cancer Navigator for personalised guidance on treatment options, specialists, and clinical trials. We're here to help you feel more confident in your choices.
References
https://www.ejoncologynursing.com/article/S1462-3889(25)00154-1/
https://obgyn.onlinelibrary.wiley.com/doi/10.1111/1471-0528.70095
https://www.physiotherapyjournal.com/article/S0031-9406(25)00376-1/fulltext
https://wjmh.org/DOIx.php?id=10.5534/wjmh.250144
https://cancer.jmir.org/2025/1/e70176
https://cancer.jmir.org/2025/1/e68426
https://ojs.aaai.org/index.php/AIES/article/download/36646/38784