Designing Care into the System: Age, Universal Design, and the Limits of AI Empathy

Artificial intelligence systems are increasingly deployed in contexts that serve older adults, yet they are frequently evaluated on the basis of linguistic fluency rather than structural inclusion. This paper argues that current large language model based AI systems fail a fundamental Universal Design test when applied to ageing populations. Two structural failures are identified: the systematic underrepresentation of older adults in training data, and interaction design assumptions calibrated to digitally fluent younger users. These failures produce simulated empathy, a form of linguistically competent but structurally misaligned responsiveness. Drawing on the Timeless pilot, a voice-first AI lifestyle ecosystem designed for older adults, the paper demonstrates how Universal Design principles can be embedded at the level of AI infrastructure through affordance recalibration, pacing modulation, and cross-domain consolidation. A three-part evaluative framework is proposed, centred on representational inclusion, affordance inclusion, and institutional inclusion, to assess AI systems against Universal Design standards. The framework positions AI accessibility as a structural design imperative rather than a post-hoc remediation.