Prepare for a visit
Bring clearer questions to a conversation with your care team.
Start the patient pathHealthcare AI with DR GPT™
From reactive medicine to proactive medicine.
Our mission is to help people understand and evaluate AI that supports prevention, earlier recognition of illness, and better care. Led by Harvey Castro, MD, MBA, known as DR GPT™, ChatGPT Health connects evidence, practical education, and responsible implementation across healthcare.
Independent of OpenAI. Educational information, with human judgment at the center of healthcare decisions.
Explore healthcare AI
Explore healthcare AIUse the governance guide
Use the governance guideClinical evaluation
Neural Networks
Current clinical data
.svg)
Human oversight
What would you like to do?
Evaluate an AI tool
Separate a product claim from the evidence that supports it.
Start the evidence pathPlan a responsible pilot
Define the task, human oversight and conditions for stopping.
Start the team pathExplore healthcare AI
Explore the questions that matter to you
Browse 12 connected topics and practical guides, from prevention and patient questions to clinical workflows and responsible adoption.
Start with your role
Find the guidance you need
Patients and caregivers
Understand where healthcare AI has helped, where it has failed, what questions to ask, and when to involve a qualified clinician.
Clinicians and researchers
Evaluate evidence, intended use, local validity, subgroup performance, workflow fit, and meaningful patient outcomes.
Health system leaders
Build accountable governance with risk tiers, approval gates, monitoring, incident response, and authority to pause unsafe systems.
Evidence before hype
Improve patient care by learning from both success and failure
Healthcare AI can expand access, reduce burden and improve detection. It can also miss cases, worsen unequal access, compromise privacy or lead people to trust incorrect output. We examine the evidence for each use and its consequences for patients.
What the Record ShowsRead sourced cases of benefit, harm, uncertainty, and the governance lesson from each.AI Healthcare GovernanceUse a risk based lifecycle for evaluation, deployment, monitoring, incident response, and retirement.About Harvey Castro, MD, MBAMeet the physician and healthcare AI advisor behind this independent educational initiative.
Independent disclosure: ChatGPT Health is physician led and is not affiliated with OpenAI or the ChatGPT Health consumer product. This site provides education, not medical advice.
Governance & Accountability
Start with the intended use, the evidence required, and the person accountable for the decision. Our governance guide connects these questions to validation, monitoring, and escalation.
Human oversight
Human oversight should match the consequence of the task. High impact clinical decisions need qualified review, authority to disagree, usable evidence, and a tested escalation path.
Accountability Frameworks
Who is responsible when AI fails? Establishing clear liability protocols for developers, institutions, and clinicians is our priority.
Bias & Hallucination
LLMs can produce confident factual errors. Test their outputs against verified clinical information before deployment, and investigate patterns that the underlying data do not support.
Auditability
Record intended use, model and data versions, material inputs and outputs, human actions, overrides, incidents, and changes while respecting privacy and security.
AI literacy
Teach people when to trust, question, or stop
Safe use requires judgment about the task and its consequences. These learning paths help patients, clinicians and leaders check important claims and seek qualified help when an error could cause harm.
For patients
Use AI to prepare questions and understand options, while keeping diagnosis and treatment decisions with qualified clinicians.
For clinicians
Check provenance, fit to the patient, missing context, and whether the output changes a consequential decision.
For leaders
Train teams on approved uses, prohibited uses, escalation paths, incident reporting, and performance monitoring.
Dr. Harvey Castro, MD, MBA
Strategic Advisor & AI Healthcare Expert
The physician behind this site
Harvey Castro, MD, MBA, is the physician behind ChatGPT Health. Explore the site's mission, published work, and approach to healthcare AI education, or contact him about advisory and speaking work.
Contact Dr. Castro
Contact Dr. CastroStandards translated into practice
Use the right framework for the right decision
WHO, NIST, FDA, AMA, and the EU AI Act address different parts of responsible health AI. We map their principles to concrete owners, evidence, controls, and review gates without treating any one framework as universal.
Clinical safety
Define intended use, human review, failure modes, escalation, and evidence needed before deployment.
Governance
Assign accountable owners, document decisions, monitor performance, and maintain an auditable incident process.
Regulatory fit
Determine which rules apply by jurisdiction, product function, risk, and whether the system is regulated medical software.
Resources & Books
Books on healthcare AI and responsible practice.
View all resources
View all resources
Books
January 2023
ChatGPT and Healthcare: The Key to New Future of Medicine
Books
February 2023
ChatGPT and Healthcare: Unlocking The Potential Of Patient Empowerment
Books
Coming Soon
ChatGPT AI and Healthcare: The Key to the New Future of Medicine
Clear answers
Frequently asked questions about AI in healthcare
Is ChatGPT Health affiliated with OpenAI?
No. ChatGPT Health is an independent education and governance site founded and led by Harvey Castro, MD, MBA. It is not affiliated with OpenAI or the ChatGPT Health consumer product.
Does this site rank foundation models as good or bad?
No. We evaluate evidence, intended use, workflow, patient impact, and governance controls. The same model can present different risks in different contexts.
Can AI improve patient care?
Yes, in well defined settings AI can support detection, access, documentation, and clinical workflow. Benefits must be measured against false positives, missed cases, inequity, burden, privacy risk, and other downstream effects.
Can healthcare AI cause harm?
Yes. Published examples include weak performance after external validation, biased allocation caused by poor proxy choices, and increased recalls or procedures without clear clinical benefit. That is why local validation and ongoing monitoring matter.
What does meaningful human oversight require?
The reviewer needs appropriate expertise, enough time and information to challenge the output, authority to override it, and a clear escalation path. A required click is not meaningful oversight.
Is this medical or legal advice?
No. This site provides general education. Healthcare decisions require qualified clinical judgment, and governance decisions require review for the specific product, jurisdiction, and organization.