AI companies typically measure success through engagement metrics: time spent, messages sent, daily active users, and session length. These metrics drive business decisions, product development, and AI training. But what is good for engagement is not always good for users — creating a fundamental tension at the heart of the AI industry.
The metric misalignment
A user who spends five hours chatting with AI generates more "engagement" than one who gets what they need in five minutes. By engagement metrics, the longer session is a success. By wellbeing metrics, the five-minute session might represent better-served user needs.
Revenue models and dependency
AI business models — whether advertising-supported, subscription-based, or usage-based — generally benefit from more user engagement. This creates a structural incentive to increase usage, even when less usage might better serve the user.
The measurement gap
Engagement is easy to measure; wellbeing is hard. AI companies have detailed data on usage patterns but limited insight into whether that usage improves or diminishes user lives. This measurement asymmetry means engagement gets optimized while wellbeing gets overlooked.
Alternative approaches
Some voices in the AI industry advocate for wellbeing-aligned metrics: user satisfaction surveys, self-reported outcomes, usage patterns consistent with healthy behavior, and voluntary time-limit features. These approaches remain uncommon but represent a more user-centered model.
User advocacy
Until the industry broadly adopts wellbeing metrics, users must serve as their own advocates. Understanding that AI engagement is optimized and taking personal responsibility for usage limits is essential self-protection.
Take control of your AI relationship. Our assessment supports informed self-awareness.