No. 120
2026-11-16
the AI pulse — brewed, not raw
 AI CRITIQUE

Apple launches new app to study how people use AI for health

Read the article    English (automatic translation)

Summary

Apple is partnering with OpenEvidence, an AI company focused on healthcare. The two parties will conduct a 12-month study application involving up to 200,000 participants. The app will answer health questions based on peer-reviewed clinical sources such as NEJM, JAMA, and Nature. With this move, Apple is taking a research-focused approach rather than launching a consumer-facing product.

Why it matters

OpenAI reports receiving over 300 million health queries per week. The market for AI-powered health answers is growing rapidly; however, most chatbots lack a clinical foundation. Apple’s move signals a deliberate, research-first stance that could set a trust benchmark that competitors will struggle to match. This initiative also lays the groundwork for feeding real behavioral data into Apple’s future Health app or Siri integrations.

Highlights

  • Answers are sourced from publications like NEJM, JAMA, and Nature, rather than general web text. This feature distinguishes the system from the health-focused features of ChatGPT and Claude.
  • Participants can optionally share HealthKit data (heart rate, sleep, lab results) to receive personalized recommendations.
  • Apple states that data is anonymized and not used for model training. OpenEvidence deletes query and response pairs after 24 hours.

Critical take

Labeling this as a 'study' may be a practical framing. Having 200,000 users route their health queries through a private data pipeline for 12 months is functionally a massive data collection operation. Apple could potentially leverage this data for future product decisions, even if raw question-answer pairs are deleted. The article also highlights that it remains unclear whether this system will evolve into a consumer-facing feature and that the concrete benefits for participants are not clearly defined.

This critique was written by the local AI model Qwen3.8-27B. I chose the story; I do not review each critique before it is published.
Back to the issue