How to Use AI for Your Health – Insights from UCSF’s Chief of Medicine

Overview

The Silicon Valley Girl podcast brings Dr. Robert Wachter, chair of the Department of Medicine at UCSF, into a candid conversation about the realistic ways AI can be woven into everyday health management.

From uploading six years of lab results and 23andMe data to an AI assistant, to deciding when a chatbot’s answer should be trusted, the episode cuts through hype and offers a pragmatic roadmap for patients and families. Wachter stresses that AI is already reliable for routine tasks—like cholesterol or urinary‑tract‑infection guidance—but remains far from replacing the human touch needed for serious diagnoses, emotional support, or complex treatment decisions.

Listeners walk away with a clearer picture of what AI can do today, where it still falls short, and how to ask the right questions to get useful, safe advice.

Key Takeaways

AI can reliably handle many low‑complexity medical queries, but it should not replace doctors for serious or emotional decisions.
Providing an AI with a comprehensive health record improves answer quality, yet too much uncurated data can introduce confusion.
Patients must frame questions like clinicians—considering both common possibilities and worst‑case “red‑flag” scenarios.
The hype around AI curing cancer is overblown; current breakthroughs lie in early detection and precision‑medicine tools.
Building a personal AI health coach is feasible using existing services like Perplexity and large‑language models.
Fact‑checking health trends with AI requires cross‑referencing reputable sources, not blind reliance on a single answer.
Within the next five years AI will likely transform specific disease areas such as diabetes management, obesity (GLP‑1 therapies), and certain cancers.

AI’s Sweet Spot: Routine Care vs. Complex Decisions

Wachter points out that “AI can treat cholesterol, blood pressure, urinary‑tract infections… that are really pretty straightforward and pretty algorithmic. ” In his hospital rounds, he still meets families grappling with life‑changing choices, like whether to continue aggressive treatment for metastatic cancer.

He explains that an AI “won’t be able to look you in the eye, hold your hand, and hug you” during those moments, underscoring the irreplaceable human element in high‑stakes care. This contrast helps listeners understand where to lean on AI for convenience and where to defer to a physician’s judgment.

Feeding Your Health Data to an AI: Benefits and Pitfalls

The host describes her own experiment: uploading six years of lab results and raw 23andMe data into Perplexity, then using the model to generate a PDF summary for her next cardiology visit. Wachter agrees that “the more they know about you, the better customized answers they can give,” but warns that “if you have too much information, they start emitting facts just because it’s too much context.

” He cites a chief medical officer who notes that medical records can be “600 pages long,” making it impossible for a doctor to read them all in a short consult. AI’s ability to synthesize and summarize that volume is a real advantage, provided the tool can reconcile conflicting entries.

Ask Like a Doctor: Framing Questions to Avoid Confirmation Bias

When patients ask a chatbot, “I have a rash, could it be X? ” the model may be biased toward confirming the premise.

Wachter advises a two‑pronged approach: “list the most likely possibilities and then ask for the worst things I could be missing. ” He illustrates this with a back‑pain example, prompting the AI to consider both benign muscular strain and red‑flag conditions like infection or tumor.

This mirrors a clinician’s mental checklist of “common vs. serious” and helps the AI produce a balanced differential diagnosis rather than a narrow affirmation.

Practical Applications

Gather your recent labs, medication list, and any direct‑to‑consumer genetic reports into a single digital folder.
Use a trusted LLM (e.g., Perplexity, OpenAI’s GPT‑4) to create a concise health summary you can share with any new specialist.
When querying an AI, structure the prompt with two parts: “What are the most likely causes?” and “What are the serious red‑flag conditions I must rule out?”
Reserve AI for follow‑up questions on medication side effects, lifestyle tweaks, or interpreting standard test results, but schedule an in‑person visit for any new, severe, or emotionally charged concerns.
Cross‑check any AI‑generated health advice with reputable sources—peer‑reviewed journals, official guideline websites, or a trusted physician—before acting on it.

Final Thoughts

This episode makes clear that AI is not a silver bullet for healthcare, but a powerful adjunct that can democratize access to reliable information, streamline routine management, and free clinicians to focus on the human side of medicine. As the technology matures, the balance will shift toward more autonomous care for low‑complexity conditions while preserving the doctor’s role for the nuanced, emotional, and high‑risk decisions that define true medical practice.

Why This Matters

Understanding where AI adds value and where it falls short helps patients harness technology responsibly, reduces unnecessary doctor visits, and accelerates the adoption of AI tools that genuinely improve health outcomes.

Source

Podcast: Silicon Valley Girl

Guest: Dr. Robert Wachter

Channel: Silicon Valley Girl

Published: October 23, 2026