ChatGPT Health Feature Tiers Medical Advice by Subscription

What happened

OpenAI has begun rolling out a feature called Health in ChatGPT to users in the United States, enabling the platform to connect with Apple Health, medical records and third-party wellness applications. The integration allows ChatGPT to draw on personal health data when responding to medical and wellbeing queries — a significant expansion of the assistant's role in everyday healthcare decisions.

The launch comes with a notable structural divide: the more capable GPT-5.6 Sol model, which powers health responses for paying subscribers, is withheld from free-tier users, who instead receive answers generated by the lighter GPT-5.5 Instant model. In practical terms, this means the quality of health guidance a person receives is directly tied to whether they hold a paid subscription.

Why it matters

Health is among the highest-stakes domains in which any service can operate. When more than 300 million people are already turning to ChatGPT each week with health questions — before this feature even launched — the platform has effectively become a front-line triage layer for a vast portion of the population. Tiering the quality of that advice by payment status introduces a behavioural and ethical dynamic that CX and service-design professionals cannot ignore: the experience a user receives is no longer shaped solely by their need or their query, but by their ability to pay.

From a behavioural-economics perspective, this is a textbook application of quality-gated freemium design — a model well understood in software, but far more consequential when the "product" is medical guidance. Users on the free tier may not realise they are receiving a materially inferior response; the interface looks identical. That invisible asymmetry is precisely what makes this design choice so significant for anyone thinking about trust, informed consent and the ethics of AI-mediated service.

By the numbers

• 300 million+ people ask ChatGPT health-related questions every week, according to OpenAI's own figures cited by The Decoder.
• Two distinct models underpin the new health feature: GPT-5.6 Sol for paid subscribers and GPT-5.5 Instant for free users.

The Renascence take

Most of the debate around this launch will focus on AI safety and data privacy. Those are legitimate concerns — but the more quietly disruptive issue is what this moment reveals about the future of tiered trust in AI-assisted services.

The real risk here is not that the free model gives bad advice — it is that users cannot tell the difference. When a service withholds quality invisibly, it erodes the foundational CX principle of informed experience: people deserve to know what they are actually receiving. A customer-obsessed operator in this space would make the capability gap explicit at the point of query, not bury it in a pricing page. More broadly, any organisation considering AI-assisted service tiers should ask a harder question first: in which domains is a two-speed experience ethically defensible, and in which does it quietly transfer risk onto the very customers you claim to serve?