Abstract
AI-enabled coaching applications are increasingly used to guide endurance training, yet athletes often worry that algorithmic plans may feel controlling or stressful. Using 90 English-language Apple App Store reviews posted in 2025 for Garmin Connect, Runna, and TrainingPeaks, I examine how users describe AI coaching in practice. Reviews were manually coded on six constructs: competence support, autonomy support, controlling style, trust, technostress, and adherence intent (0–2 scale; strong signals used for comparison). Across the sample, competence support dominates and explicit control is rare. Adherence intent is strongest where competence support is paired with low friction and reliable performance; technostress and trust breakdowns cluster where users report syncing failures and governance friction (e.g., subscriptions and feature lockouts). Overall, the evidence suggests trust links competence support to continued use, while stress and perceived control stem mainly from platform governance and reliability rather than coaching logic itself.