The VERIFY Framework
Six steps to evaluate any health information you get from AI. Click through each step to learn what to look for, why it matters, and where to go deeper.
Verify the Source
What to do
Ask the AI to show you the sources it used to generate its response. If it can't cite specific sources, that's your first warning sign.
Click through to the websites it cites. Read the original content and compare it to what the AI told you. Does the source actually say what the AI claims it says?
Check website credibility: Who wrote it? When was it published? Is the domain reputable (.edu, .gov, established medical institutions)? Is there an author with real credentials?
Evaluate Confidence
What to do
A polished, fully confident answer with zero hedging is not proof of accuracy. Tone barely tracks accuracy — so uniform confidence is your default warning sign, not reassurance. This is the "confidence gap."
An explicit hedge — "I'm not certain," "this may vary" — is an honest signal, not a flaw. Chatbots do sometimes hedge; when one does, take it seriously and verify that specific point yourself before acting on it.
As a low-cost tiebreaker — not a verification step — you can ask the AI how confident it is. Treat its self-rating with caution: research shows models tend to agree with whatever a question assumes, so the AI's own confidence can never substitute for checking a real source.
Learn More
When AI Gets It Wrong: Addressing AI Hallucinations and Bias — MIT Sloan Teaching & Learning Technologies AI hallucinations can't be stopped — but these techniques can limit their damage — Nature (news feature, 2025) Survey of Hallucination in Natural Language Generation — ACM Computing Surveys (Ji et al., 2023)Recognize Limitations
What to do
AI isn't perfect. Accept that as a starting point, not an afterthought.
Some topics are too broad or have limited research behind them — these are exactly where AI is most likely to get things wrong. If the topic is niche, emerging, or under-studied, be extra cautious.
For high-stakes decisions — anything that could significantly affect your health, career, or wellbeing — it's always better to conduct your own research from primary sources.
Think about the cost of being wrong. If an incorrect answer would cause real damage, that's your signal to verify independently.
Investigate Bias
What to do
Check whether the AI's response contains bias. AI models are trained on existing data — and existing data reflects existing inequities. The answer you get may not apply equally to everyone.
AI remembers the facts you give it within a conversation. Make sure it's not taking your inputs out of context and feeding you back misinformation that could harm you.
If you detect bias in a response, conduct your own research targeting the factors that are most relevant to your specific situation — your background, your health history, your demographics.
Bias is hard to spot, especially when the AI sounds confident. Even small, seemingly unimportant biases can compound into bad advice. When in doubt, find a reliable source on the web.
Flag Red Flags
What to do
If any piece of information seems incorrect or you're unsure about its validity, it's better to disregard the entire response and find a different source. Don't try to salvage partially wrong answers.
AI builds its responses on its own previous statements. One red flag can contaminate everything that follows — even if the rest of the response looks correct, it may be built on a flawed foundation.
Red flags in health-related responses are especially dangerous. Wrong medical information doesn't just waste your time — it can lead to real harm. Treat health red flags as non-negotiable stop signs.
Your Health, Your Call
What to do
AI is supposed to be a tool — not a decision-maker. Use it to start your research, not to end it. Don't outsource your thinking.
If a response doesn't match what you expect to see — or something feels wrong — trust that instinct. Find a reliable source and conduct your own research.
Most of the time, conducting your own research leads to safer and more accurate results. AI can help you get started, but the final call on your health is always yours.