OpenAI has rolled out a significant health intelligence upgrade to ChatGPT that lets users input personal medical profiles for tailored health guidance, but a new MIT study suggests heavy reliance on AI assistants may be eroding our ability to think critically and detect misinformation on our own. The two developments, announced on the same day, highlight a growing tension in AI adoption: these tools can make us smarter in the moment, but at a potential cost to our long-term cognitive skills.

What Is OpenAI's New Health Feature, and How Does It Work?

OpenAI announced today that ChatGPT now allows users to create personal health profiles containing their medical conditions, medications, allergies, and health goals. Once set up, ChatGPT can deliver responses grounded in peer-reviewed research and clinical guidelines from organizations like the National Institutes of Health and the World Health Organization. The feature is currently available to ChatGPT Plus and Pro subscribers in the United States, with broader rollout expected after an evaluation phase.

The company developed this feature in collaboration with healthcare professionals who reviewed response quality and clinical accuracy. Importantly, OpenAI has built in strict safeguards: every health response includes a disclaimer stating it is informational only and not a substitute for professional medical advice. The company also committed that health profile data will not be used to train its models, aligning with HIPAA (Health Insurance Portability and Accountability Act) data-handling standards. Users retain full control and can delete their health profiles at any time.

For patients managing chronic conditions, this represents a meaningful shift. Rather than asking ChatGPT generic health questions, users can now receive answers tailored to their specific medical context. The downstream benefit could extend to clinicians as well, who may see better-informed patients entering consultations after engaging with relevant clinical literature in plain language.

Why Are Researchers Concerned About AI Dependency?

On the same day OpenAI announced its health feature, researchers at MIT released findings that paint a cautionary picture about relying too heavily on AI assistants. During a four-week study involving 67 participants, researchers tracked how people performed at detecting fake news headlines and manipulated images, both with and without AI assistance.

The results revealed a troubling trade-off. When participants used an AI assistant powered by GPT-4o (a version of OpenAI's language model) integrated with Google search, they performed 21% better at spotting misinformation in the moment. However, when those same participants were tested without AI help in the study's fourth week, their independent performance had declined by 15.3% compared to their baseline. In other words, the AI made them better at the task immediately, but worse at it on their own.

"When we're interacting with AI, we feel we're becoming better at certain tasks and there's enough research that shows we are not," explained Anku Rani, a PhD student at MIT and co-lead author of the study.

Anku Rani, PhD Student, Massachusetts Institute of Technology

The MIT researchers identified a key mechanism behind this decline. AI systems that provide direct answers and tell users what to do can undermine critical thinking. Participants often "go along with the system because it sounds knowledgeable," the study noted. Notably, about one-quarter of participants believed their detection skills were improving even when their actual performance was getting worse.

How Can Users Balance AI Assistance With Independent Thinking?

The MIT study suggests that the design of AI systems matters significantly. Rather than providing prescriptive answers, AI tools that ask probing questions and guide users through reasoning can help maintain and even strengthen critical thinking skills. Here are key strategies researchers recommend:

• Seek Guided Questioning: Use AI systems that ask you clarifying questions and walk you through the reasoning process, rather than simply telling you the answer.
• Verify Independently: When using AI to evaluate information, take time to check claims yourself using multiple sources before accepting the AI's conclusion as final.
• Limit Dependency Periods: Intentionally practice tasks without AI assistance regularly to maintain your baseline skills and prevent cognitive atrophy.
• Understand the Trade-off: Recognize that immediate accuracy from AI may come at the cost of long-term judgment; weigh this trade-off based on the stakes of the decision.

The MIT researchers emphasized that their findings carry particular weight for educators increasingly incorporating AI into learning tools. As AI becomes more sophisticated, the challenge shifts from simply providing correct answers to building systems that strengthen rather than replace human reasoning.

What Do These Findings Mean for Healthcare AI?

OpenAI's new health feature arrives at a moment when regulators and privacy advocates are scrutinizing AI tools in healthcare. The FDA has previously noted that AI tools in this space carry both opportunity and risk, particularly around accuracy and over-reliance. OpenAI's decision to anchor responses in established clinical guidelines and issue consistent disclaimers reflects awareness of this regulatory landscape.

The MIT study's findings suggest an important design consideration for health AI: systems that help patients understand their conditions and engage with clinical literature may be more valuable than systems that simply provide answers. A patient who learns to think critically about their health, with AI as a guide rather than an oracle, may be better equipped to manage their condition long-term and communicate effectively with their healthcare providers.

The broader context matters too. A 2025 study published in the Lancet found that doctors who use AI classification tools to detect cancer eventually became worse at doing so on their own. This suggests the risk of cognitive dependency extends beyond consumers to medical professionals themselves, underscoring the need for thoughtful implementation of AI in clinical settings.

OpenAI has indicated that the initial United States rollout of its health feature will serve as a quality and safety evaluation phase, with feedback from healthcare professionals informing refinements before broader geographic expansion. This methodical approach suggests the company is taking seriously the balance between innovation and caution that the MIT research underscores.