<p>The COVID-19 pandemic has substantially disrupted daily life across multiple domains worldwide. Research on well-being impacts has yielded mixed findings, with some studies documenting recovery patterns while others report persistent declines, particularly among certain demographic groups. Understanding long-term trajectories and identifying vulnerable populations remains crucial for informing recovery policies. This study aims to examine temporal changes in multidimensional well-being across sociodemographic groups and their associations with pandemic control behaviors in Taiwan from the initial outbreak to the official end of the pandemic. Using longitudinal data from a probability-based online panel survey (<i>N</i> = 1,744), we analyzed changes in seven well-being domains across four time points (2020–2023). Mixed-effects models examined how sociodemographic factors influenced well-being trajectories, while effect sizes assessed the magnitude of changes over time. Results revealed that despite initial rebounds in some indicators, most well-being domains demonstrated a consistent decline throughout the pandemic period, although effect sizes were small (d = -0.04 to -0.21). Standard deviations increased significantly across domains, indicating widening disparities between individuals. Men, younger adults, lower-income individuals, and unmarried persons emerged as the most vulnerable groups, experiencing greater well-being deterioration compared to their counterparts. These findings challenge assumptions of automatic recovery and highlight the need for sustained, targeted interventions that address both the cumulative effects of prolonged stress and the differential needs of at-risk populations during large-scale crises.</p>