Risks of AI in healthcare include inaccurate outputs, uneven performance across populations, inappropriate data use, and over-reliance on a tool. A consumer wellness app, a general chatbot, and a regulated medical device are not interchangeable.
Accuracy and intended use
Check what the particular tool is designed and authorized to do. A fluent symptom explanation is not a diagnosis. A product can work well on one task and fail on another. Do not delay necessary care because a generated answer or readiness score is reassuring.

Bias and evidence
Ask which populations and settings were included in validation and where performance is uncertain. Do not assume that a result from one group applies to everyone. Bias can arise from the data, model, workflow, and how a tool is used.
Privacy and access
- Identify what information is collected and why.
- Read the current policy for storage, sharing, training use, and deletion.
- Check account access and export controls.
- Avoid unnecessary identifying details in general-purpose services.
Do not claim that every health app sells data or that anonymization removes every privacy risk. Policies and protections differ by service and context.
Oversight that can actually work
A reviewer needs relevant expertise, access to the evidence, and the ability to correct or reject output. Human approval without meaningful inspection is not a guarantee. Record what remains uncertain and how concerns can be escalated.
Illustrative failure
A fictional user receives a reassuring skin-image assessment and delays a medical appointment. The lesson is that an app output should not replace assessment of a concerning symptom. No numerical probability or clinical outcome is attributed to a real product or patient here.
Questions before using a tool
Help me review this health app's documentation: [text].
Separate intended use, evidence, limitations, data practices, and missing information.
Do not infer regulatory approval or recommend a diagnosis.
List questions for the provider or a qualified healthcare professional.The wellness-tool guide describes categories; the wellness overview explains organizational uses.
Sources and editorial approach
WHO guidance on large multimodal models for health. Reviewed November 4, 2026. Examples and prompts are proposed editorial templates, not first-hand tests or measured performance results.







2 Comments
Comments are closed.