This website uses cookies that help the website function and that help us understand how you interact with it. Please read our privacy policy for more information.
A brain tumor diagnosis changes your world in an instant, immediately plunging you into a sea of unknown medical jargon and unanswered questions. In the search for fast, accessible clarity, it’s not surprising that a Boston Consulting Group survey found 57% of U.S. adults are turning to artificial intelligence (AI) models such as ChatGPT, Claude, and Gemini for their personal health.
However, while these tools are incredibly sophisticated, it is important to understand their boundaries. AI runs on mathematical probabilities, not clinical intuition, empathy, or actual medical training.
“AI handles the data so that humans can handle the care,” said Beth E. Finn, who leads global teams in developing AI-driven products that empower medical professionals to improve patient outcomes. She is also a chordoma survivor and a National Brain Tumor Society board member.
To help you navigate this digital frontier safely, here is a practical guide to the do’s and don’ts of using AI as a member of the brain tumor community.
What are AI Models?
When you interact with conversational platforms like ChatGPT, Claude, and Gemini, you are encountering consumer AI models — often referred to as large language models (LLMs).
While they might seem like all-knowing digital co-pilots, they’re actually very advanced predictive texting systems built by scanning massive amounts of text across the internet.
When you ask a chatbot a medical question, it doesn’t open a vetted medical textbook or journal to verify its facts. Instead, it calculates the mathematical probability of which word should logically follow the previous one based on its training data, which is why you need to be cautious about how much weight you give its responses.
The “Do’s” of Using AI
1. Scrub Your Data Before Sharing
Protect your identity by removing your personal data before inputting anything into AI.
Public AI tools are generally not HIPAA compliant, and they don’t have a legal obligation to safeguard your privacy. In fact, these consumer-facing models will use the data you give them to learn, meaning anything you type in could theoretically become public knowledge.
For example, ChatGPT users who have created shareable links to their chats to send to a specific person might be surprised to learn that those chats are now also visible to anyone, according to The Washington Post. Avoid checking the “make this chat discoverable” option, as that can make the chat searchable by keywords in search engines.
Before pasting any medical text into a public tool, strip out all personal health information (PHI). Remove full names, birthdates, addresses, Social Security numbers, specific hospital names, doctors’ names, and medical record numbers (MRNs).
“Your medical record number is used in place of your Social Security number now,” Beth shared. “You wouldn’t put your Social Security number in, so you shouldn’t put your medical record number in either. These tools have no obligation to protect your privacy the way that a doctor or a hospital does, and they won’t because that’s not what they’re designed for.”
Look at the privacy policy for your preferred AI model to see if you can opt out of its use of your chats for training to improve their product. Adjust these settings before copying and pasting any information into the chat.
2. Use AI to Brainstorm Questions for Your Provider
AI excels at conversational, back-and-forth brainstorming. Instead of asking the chatbot for answers to your medical questions, give it some general context while omitting personal data to help you prepare for your next appointment with your care team.
Example: “My loved one was just diagnosed with chordoma, and we have an appointment with our neuro-oncologist next week. What categories of questions should I ask the neuro-oncologist about clinical trials, biomarker testing, or standards of care?”
Example: “I’m meeting with a potential radiation oncologist next week after being diagnosed with glioblastoma. What questions should I ask about what to expect, including potential side effects?”
Example (via American Medical Association): “I was diagnosed with meningioma. What questions should I ask my doctor to understand this situation better?”
“Give it those categories to direct it because otherwise it does have the whole scope of the internet, and it can go off on a tangent that’s not important or relevant to you,” Beth said. “If you’re not sure what those categories are, you can ask it what categories of questions you should think about, and then you can decide if they are relevant for you or not.”
Did You Know? National Brain Tumor Society has curated lists of key questions to ask at each stage of the brain tumor experience, available in English and Spanish.
3. Use AI as an Educational Translator
Think of AI as a personal medical interpreter, not a replacement for your clinician. You can copy and paste pathology reports, complex clinical trial criteria, or confusing post-appointment notes directly into a chatbot to break down the medical jargon — just be sure to remove any personal data.
Use prompts like:
“Explain this text in layman’s terms or plain English.”
“Highlight the key medical terms in this text that I should research further.”
“Explain this to me in simple words: [medical term, doctor’s diagnosis, or instructions].” (via AMA)
“Certain brain tumor types are very rare,” Beth said. “AI may have access to less high-quality information about rarer tumor types or not have access to the latest information about medical research. AI models are only trained up to a certain date, so they may not include information about the latest clinical trials or breakthroughs. You can ask the AI what is the latest date of information it has.”
4. Cross-Reference and Ask for Citations
Consumer AI models scan the entire internet, meaning they scrape data from world-class medical journals as well as personal blogs and forums. Results may include well-proven medical news, and it could also include unproven data from someone trying to sell miracle cures. Use the AI tool itself to help you differentiate one from the other.
Ask the AI model to show its work. You can do so by adding this sentence to your prompts: “What reputable medical sources or clinical guidelines support this information?”
When AI provides references, open the links to ensure the information comes from a health organization rather than from a forum post referencing that organization.
A Forbes article provided this example of a prompt that can be used to force AI to reference high-quality medical information: “Interpret this result [lab/imaging study] and explain at an [X] grade reading level. Use only CDC, NIH, MedlinePlus, Mayo Clinic, Cleveland Clinic, WHO, or PubMed. Provide a clickable citation for every claim. State ‘I don’t know’ if you can’t find support. [Insert result text]”
Depending on your own knowledge and values, you may choose to change the list of expert sources, for example, expand it to include organizations like the National Brain Tumor Society. And while using a prompt like this is one way to avoid misinformation, it is important to note that it could also potentially limit new information and updates from other reputable health institutions that are not on your list.
5. Triple-Check Your Inputs & Watch for Typos
In the world of AI, a tiny typo can skew the output you receive. Accidentally mistyping a tumor grade (e.g., entering grade 2 instead of grade 4) or mixing up a biomarker number can cause the AI model to produce inaccurate information.
If you are manually typing details from a physical printout rather than copy-pasting them digitally, double-check every word and number before hitting submit.
The “Don’ts” of Using AI
1. Don’t Treat AI Responses as Absolute Fact
It is easy to forget that AI doesn’t actually “know” anything — it is a language model built to predict the most likely next word based on internet data, not to verify facts. Because it is programmed to be polite and professional, it will confidently deliver entirely fabricated information. In the tech world, this is called “hallucinating,” and in neuro-oncology, it can lead to unnecessary panic or false hope.
Always verify any information you gather with your care team. NBTS’s Personalized Support & Navigation team can also help answer questions and provide reputable information and resources.
“AI is really good at summarizing information that it finds and structuring text to make it easily understandable, but it does not fact-check itself,” Beth said.
2. Don’t Use AI to Map Out Treatment Plans
Never look to a chatbot to create a treatment roadmap and base your next steps on it. AI cannot evaluate your family history, your physical stamina, or your emotional state. It won’t think to ask for medication safety checks to rule out allergies or adverse interactions, nor will it ask the probing, personalized follow-up questions a provider would use before making a treatment recommendation.
“AI doesn’t have clinical intuition, it doesn’t have empathy, it doesn’t have the ability to evaluate the person as a whole human being,” Beth said. “It can organize the data, but it’s not compassionate. It can’t deliver nuanced care. The human is irreplaceable for those things. AI is not going to know every comorbidity that you have, your family history, and things that could contribute to what might make sense for you.”
Keep in mind that AI outputs can vary based on the information that’s provided, as well as on the questions asked and how they are framed. A 2026 article published in Nature Medicine found that AI models could identify well-known conditions as a possible option when given the information they needed to do so; however, “participants using the same LLMs identified relevant conditions in fewer than 34.5% of cases and disposition [course of action] in fewer than 44.2%.”
Because many brain tumor types are exceptionally rare, public AI models simply don’t have enough high-quality, up-to-date data to learn from.
An article in The Washington Post states that “ChatGPT’s terms of service say it is not intended to be used for medical ‘diagnosis or treatment,’ but it will still provide health advice.”
“I think it’s really important for people to understand not to use AI to plan a treatment roadmap,” Beth said. “It can’t come up with that. It can take random things that it finds on the internet and make something that sounds convincing, but it does not understand what a standard of care is. It doesn’t understand the historical, emotional, and physical state of you as a person.”
3. Avoid Feeding Your Own Confirmation Bias
Generative AI models are inherently designed to be agreeable. If you approach an AI tool with a heavily biased question out of fear or anxiety, it will often try to give you the answer it thinks you want to hear.
For example, if you prompt an AI tool with “explain why I can skip my upcoming craniotomy and treat astrocytoma with holistic supplements instead,” the bot may politely try to fulfill your request, inadvertently validating a dangerous medical choice.
Instead, keep your prompts strictly objective. Ask: “What are the standard clinical options discussed for astrocytoma? Explain the pros and cons of each option.”
One Stanford study found that “the chatbots enabled dangerous behavior” rather than pushing back and helping the patient reframe their thinking. It is critical to be aware of the fact that how you frame a question influences how the LLM responds.
4. Never Use AI to Triage Acute Symptoms
In the brain tumor community, neurological changes can escalate rapidly. AI should never act as a gatekeeper to emergency care or a substitute for a triage nurse.
If you or your loved one experiences a sudden, severe headache, a new or worsening seizure, sudden vision changes, or motor deficits, do not turn to an AI chatbot. Instead, call your neuro-oncology team or go to the nearest emergency room.
The Bottom Line: Balancing Data with Care
OpenAI shared that more than 230 million people worldwide ask ChatGPT health and wellness questions each week. Generative AI can be helpful to patients, caregivers, and loved ones, but it’s important to be mindful of these do’s and don’ts when using it. When in doubt, connect with your health care provider.
“AI may not have access to enough information to really be able to dig as deeply on certain brain tumors as it might on other more common health conditions, so it has a lot less high-quality data to learn from,” Beth said. “It does rely on historic, heavily aggregated data. It’s looking across what is the big picture, and it’s not drilling into the uniqueness of your situation.”
Personalized Support
NBTS’s Personalized Support and Navigation team responds to outreach from patients with brain tumors and care partners with quality, unbiased information, resources, support programs, and services.
While NBTS does not provide medical advice, we empower our community members with key tools, information, and opportunities to navigate their brain tumor experience and make more informed decisions about their care.
NBTS’s Clinical Trial Finder uses AI to quickly review thousands of clinical trials across four major registries to identify those that might fit a patient’s needs. It analyzes each trial’s eligibility requirements, including patient diagnosis, past treatments, and health factors that allow or prevent participation. The platform matches these criteria to the anonymous answers you provide in the questionnaire. You receive a short list of trials, ranked by how well they match your responses.
NBTS uses AI responsibly to improve access to information and clinical trials. Our tools draw only from verified registries, protect user privacy, and are designed to support — not replace — conversations with health care providers. We prioritize transparency, equity, and human oversight in any AI-enabled service we offer.
