How my personalized AI health consultant helps me answer ‘WTF’ medical questions

9 min readMay 5, 2025

As someone who’s had autoimmune and allergy issues most of my life, I’ve become somewhat accustomed to my body doing bizarre and unexpected things. But especially in the last 15 years, I’ve had a range of symptoms that my doctors have not been able to account for, and questions they haven’t been able to answer. A few examples:

Why do I have extreme hormonal fluctuations, my body expanding and shrinking like a balloon, seemingly at random?
Why do I get an occasional rash on my insole after autoimmune treatments that is a perfect oval shape??
I recently had off-the-charts high platinum levels — is this concerning, and what could be potential sources of this exposure?

Unsolved mysteries like these, however minor, contribute to a feeling of being out of control. And when my body is doing odd, seemingly random things, I crave a sense of order and reason. I want to at least understand why these things might be happening, so that if needed, I could take some sort of action.

Inspired by Jared Zimmerman’s recent article about creating a medical second opinion GPT that used his health records in its reasoning and responses, I decided to put together a quick prototype for myself. Here’s what I did and what I learned.

What I did

I used Claude for this experiment because of Claude’s (Anthropic’s) privacy policy. Claude does not train its model on your data, and if you delete a chat, the data is deleted from the system within 30 days.

Claude has a concept called ‘Projects,’ which I believe is similar to GPTs in ChatGPT. You can add instructions and upload files into the ‘Project Knowledge,’ and Claude will refer to these when answering questions.

I attached two files to the project. The first included recent blood test results from a company called Genova, which had detailed recommendations about vitamins and minerals, heavy metals, antioxidant support, and more. This is one of those holistic, pay-out-of-pocket tests. The report had a complex layout, with lots of different graphs and visuals, so I wasn’t sure if Claude would be able to use it.

The second file was a PDF that included a visual timeline of my whole health history, generated by Pictal Health (my startup). I was also curious to see if Claude could parse this PDF and use it to answer questions. Below is a very zoomed-out view of it. In brief, there are different swim lanes for symptoms, diagnoses, hospital stays, acute illnesses, life events, medications and treatments, substance use, surgeries, and tests/imaging. There’s a lot there.

I used Jared’s instructions as a starting point for my written project instructions, which included a list of current medications and doses, adverse effects of prior medications, prior diagnoses, biometric averages, body measurements, and I added some recent allergy testing results. I also added instructions for how to interpret the visual timeline document. If you’re going to do this for yourself, I recommend using Jared’s experiment as a starting point.

The whole thing took about 45 minutes (expedited because I already had compiled my health history), and I was up and running with my new health consultant Claude.

What I’ve learned so far

To my delight, Claude was able to read and provide a summary of both of my complex PDFs. And within minutes, I had answers to some of my longtime questions. In case they are of interest:

Why my hormones may be out of balance: A supplement I take in large quantities, quercetin, may be contributing to my hormone issues. Claude suggested a few tapering strategies for slowly reducing my quercetin intake. Claude also explained three other reasons why my hormones may be out of balance, given some of my specific test results and my history of histamine intolerance. And, Claude reminded me about a few supplements I had discussed with my doctor, but that I had forgotten to start.
What’s up with this random circular foot rash: This seemed to Claude to be consistent with erythema annulare centrifugum (EAC), a type of hypersensitivity reaction. It appeared to be an occasional reaction to my autoimmune treatments. Claude said, “This type of reaction is not uncommon with IVIG therapy and is generally considered mild” — which helped ease my mind.
On the high platinum levels: Claude explained how this could be making my histamine intolerance issues worse, and that there could be immune system impacts of high platinum. Claude also suggested potential sources of platinum in my environment.

I now find myself checking in with Claude about health questions every few days.

Why do I have a cooling sensation in the front of my brain when I eat certain foods? Claude: perhaps a “unique neurological response to beneficial compounds rather than a concerning symptom.”

Are there any potential interactions between my supplements and medications? Claude informed me that Magnesium may interfere with one of my main immunosuppressant medications, so I should take magnesium at another time of day; I should also be careful with probiotics, since they could cause infection while I’m on immunosuppressants.

What are highlights from my history that a neurologist may want to know? Claude pulled from my timeline and provided a succinct, bulleted summary that included diagnosis, summary of treatments, adverse treatment reactions, summary of flare-ups, and other facts from my history that may be relevant. I may show this to my neurologist, just to see what he thinks of it. Here’s an excerpt:

This type of summary would be extremely helpful to bring to a first appointment with a new doctor.

Benefits I have seen

To me, chatting with Claude about my health history has been like having a super smart holistic health coach that never forgets…anything. Apart from finally getting answers to those obscure questions that had been dogging me, here are a few other benefits I have experienced:

I have a thought partner capable of dealing with my complexity

With 5+ diagnoses, 17 current medications and supplements, and a whole lot of medical history, it’s tough for any one provider to give me advice with my whole history in mind. It’s too much for one person to hold in their brain at once (that’s why I’ve been working to visualize health histories.)

I’ve found that Claude gives me holistic answers, taking into account my whole story, including physical health and mental health. Claude also has access to an internet’s worth of information, so it (he?) was able to put a scientific name to a symptom, like my circular rash, which may be rare. Finally, Claude was able to summarize my neurological history within the context of my whole history, a feat that would be difficult and time-consuming for a human.

I no longer feel alone with my mini-health mysteries

As I mentioned, the questions I have been asking Claude are not life-threatening, urgent health issues. (For those, as you know, we should all seek immediate medical help.) They are more often loose-end type of questions that doctors haven’t been able to answer, or things that occur to me at randomly at 10pm.

Getting some answers to these questions is reassuring. There is a sense of relief in knowing I can get “good enough” answers from Claude, when I can’t get answers through traditional methods. I feel less alone. I feel like Claude has my back. As strange as that sounds.

I’m reminded of things I intended to do

Claude has consistently checked in about my test results, reminding me of areas that may need attention. For example: “I notice there are some nutritional areas that may warrant attention, particularly regarding your antioxidant needs, B vitamins, and omega-3 fatty acids.” Dang, I’ve been forgetting to take my omegas! These kinds of reminders and nudges are helping me stay on track with my doctor’s advice; I need that help, as I only meet with her every 6 months or so.

Risks

There is always a risk that Claude is giving me incorrect information. I would not start any Claude-recommended supplements or medications without talking it over with my doctor first. For any more urgent or serious issues, I would focus on working directly with the healthcare system, though I do think AI can help in hard-to-diagnose situations.

There’s also a risk of over-relying on AI for medical care, and losing some of the benefits of human interaction and touch.

There are surely many more risks — please leave thoughts in the comments.

Opportunities

Where do we go from here? Here are a few ideas, from my vantage point as a complex patient who is obsessed with helping patients tell their stories.

With a patient’s whole health history, AI can become an effective health consultant. Though it will be very cool to query our own medical records for insights, we will see limited benefits if we only work with the data that exists in digital health records today. Significant life events, symptom history, health history from childhood (for us older folks,) past exposures to toxins or molds; these things all have a bearing on our health outcomes, and in many cases this information is not well-captured in our health records. It needs to come from the patient. This is what we are working on at Pictal Health.
AI can help patients build and visualize their stories. With a combination of medical data aggregation and speech-to-text tools, we can turn the process into a seamless, empathetic conversation. I do still believe that it’s important to be able to see your health story visually, even in a world where AI can summarize complex data; I will break down the reasons in a future post.
As a holistic health coach, AI can help people meet their health goals and keep up with medications and supplements. I have seen this already in my limited interactions with Claude. It can be there for us between doctor visits, nudging us to change our behavior, and it can be an accountability partner that helps us sustain changes over time.

For complex patients like me, who manage the bulk of our day-to-day care, AI tools can serve as coach, thought partner, and extra support layer so that we feel less alone with our health. I would love to hear your thoughts, and if you try this for yourself, let me know how it goes.