Early in my career, I learned the difference between building the model and signing it.
The junior analyst who can’t sign
I spent over 15 years in corporate finance, and I worked with some brilliant junior analysts. Hand one a messy data room on Friday and by Monday you’d have a clean model, a list of anomalies and three questions nobody else had thought to ask. The good ones were fast, tireless and, every so often, wrong with total confidence.
None of them signed the deal. Not because they weren’t smart, but because signing means accountability. The person who signs knows the full picture, answers for the outcome and is allowed to say “no.”
That’s how I think about AI and your health data. If you’re like a lot of the founders and executives I talk to, you already track everything: sleep and recovery from a wearable, labs a couple of times a year, a supplement stack that keeps growing. And lately you’ve started pasting those results into a chatbot and asking it what to do.
So, can ChatGPT read lab results? In a sense, yes. AI tools can read them, explain them and organize them remarkably well. What they can’t do is sign off. AI is a brilliant junior analyst for your optimization data. It just isn’t allowed to sign anything.
I’m the founder of The Vitality Collective, not a clinician, so this isn’t medical advice. It’s the rule I use.
What AI is genuinely good at
Used as an analyst, AI can save you real time and make you a better-prepared patient:
- Decoding the jargon. Lab reports are written for clinicians. AI can explain what a marker generally measures, what a “reference range” is and why a flag on a report isn’t a diagnosis.
- Spotting trends. Give it a few years of labs and your wearable exports, and it can lay out what has moved, in which direction and around which life changes: a new role, a move, a heavy training block.
- Summarizing a study, including its limits. Ask it whether a study was done in people or animals, how many participants it had, how long it ran and what it actually measured. That’s often where an exciting headline falls apart.
- Building a sharp question list. This is the most useful job of all. A good list turns a rushed appointment into a productive one.
Where it breaks
Here’s the part that tends to get skipped.
It only knows what you tell it. It doesn’t have your full history, your medications, your family history or the thing you forgot to mention (or chose not to). The same lab value can mean different things depending on when it was drawn, what you take and what else is going on in your life. I walked through one example, testosterone timing, in Is It Your Hormones or Just Your Thirties?
It can’t examine you. No vitals, no physical exam and no follow-up question based on how you look when you answer.
It isn’t accountable. If it’s wrong, nothing happens to it. The companies that build these tools say as much. OpenAI’s usage policies, for example, don’t allow its services to be used for tailored advice that requires a license, such as medical advice, without appropriate involvement by a licensed professional.
A confident tone doesn’t make it right. The World Health Organization has pointed to documented risks of these models producing false, inaccurate, biased or incomplete statements that could harm people making health decisions. The answer always sounds fluent, and that’s the trap. A polished paragraph feels settled even when the evidence underneath it isn’t, and that includes unproven “optimization” protocols that get presented as if the science were finished.
The AI-built protocol problem
There’s a specific moment where the relationship changes. It’s when the prompt goes from “help me understand this” to “design my stack,” “build me a dosing schedule” or “write me a protocol.”
At that point the tool has quietly moved from analyst to prescriber. It’s choosing what goes into your body, how much and how often, from an incomplete file and with no one accountable for the result.
Prescription decisions belong to a licensed clinician who has your full picture, can order the right tests and can say “no” or “not yet.” Those two answers aren’t obstacles. They’re part of good care. Some of the most valuable advice in any field is the advice that talks you out of something.
A better prompt, and a privacy check
Here’s the kind of prompt I’d actually use:
“Here are my last three lab results and some notes on how I’ve felt over the same period. Explain what each marker generally measures, show me what changed and give me a list of questions to ask a licensed clinician.”
Notice what it doesn’t ask for: a diagnosis, a dose or a protocol.
Before you upload anything, run a quick privacy check. Read what the tool does with your data: whether it’s stored, whether it’s used to train models and whether you can delete it. Don’t assume HIPAA comes along for the ride. HIPAA generally doesn’t cover health information you give an app yourself. Even when a doctor’s office sends your records to an app at your request, HHS says that once the app has them, and it isn’t a HIPAA covered entity or business associate, the information is no longer protected by the HIPAA Rules. At a minimum, consider removing your name, date of birth and other identifiers first.
Analyst, not prescriber
Use AI the way a good finance team uses a junior analyst: for speed, structure and better questions. Then take the work to someone who can sign.
Prep with AI. Decide with a licensed clinician. Then re-measure, because the only way to know whether a change worked is to check. It’s the same discipline I wrote about in You Underwrite Your Executives. Who’s Underwriting Their Health?
When to talk to a clinician sooner
Don’t wait on a chatbot if your symptoms are persistent or getting worse. If you’re feeling persistently hopeless or have thoughts of harming yourself, call or text 988 now. In an emergency, call 911.
Bring better questions
If you want a head start on that question list, our free Is It Your Hormones? self-check walks you through common symptoms and gives you a short checklist of labs worth asking a licensed clinician about. It won’t diagnose you, but it will help your next appointment start in the right place.
Would you rather talk it through? Request a Private Call. Available nationwide. Some visit requirements vary by state. Medical care through The Vitality Collective is provided by independent licensed clinicians.
— Charles
Sources
- OpenAI. Usage Policies. https://openai.com/policies/usage-policies/
- World Health Organization. WHO releases AI ethics and governance guidance for large multi-modal models. January 18, 2024. https://www.who.int/news/item/18-01-2024-who-releases-ai-ethics-and-governance-guidance-for-large-multi-modal-models · Full guidance: Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models. https://www.who.int/publications/i/item/9789240084759
- U.S. Department of Health and Human Services. HIPAA FAQ 3009: Does a HIPAA covered entity that fulfills an individual’s request to transmit ePHI to an app bear liability for the app’s use or disclosure of the health information it received? https://www.hhs.gov/hipaa/for-professionals/faq/3009/does-a-hipaa-covered-entity-bear-liability.html
- 988 Suicide & Crisis Lifeline. https://988lifeline.org/
Educational only, not medical advice. I’m the founder of The Vitality Collective, not a clinician. Talk with a licensed clinician before starting any treatment. Medical care through The Vitality Collective is provided by independent licensed clinicians. Individual results vary. Available nationwide. Some visit requirements vary by state.