Explainable AI (XAI) promises to make AI decisions understandable to humans, but for whom? We conducted systematic search from Scopus, for comprehensive accessibility XAI and for agentic/LLM XAI with accessibility search. Analysis reveals a profound medicalization gap: 76 papers (98.7\%) use XAI to diagnose or study accessibility populations, while only one design explanations for users. The agentic search shows the same pattern: all 10 papers use XAI to study populations, with none designing accessible explanations for users of agentic/LLM systems. This reveals current research addresses clinicians' and developers' needs but overlooks disabled users' accessible explainability needs. We propose shifting from XAI \textit{on} populations to XAI \textit{for} users with concrete recommendations for inclusive needs assessment.

Proceedings of the CHI 2026 Workshop on Human-Centered Explainable AI (HCXAI); April 13–17, 2026; Barcelona, Spain.