<p>In high-stakes visual search tasks, such as reading a medical image or screening luggage at an airport security checkpoint, targets are often rare, leading to disproportionately high miss rates—a phenomenon known as the low-prevalence effect (LPE). While the LPE is traditionally attributed to conservative shifts in decision criterion (c), recent research on prevalence-induced concept change (PICC) suggests that low prevalence can lead to liberal decision criterion shifts in certain contexts. We propose that trial-level performance feedback is a critical mechanism governing these divergent outcomes. We conducted a meta-analysis of 62 effects from 29 publications, using robust variance estimation (RVE) to account for statistical dependencies. Our results confirm a large and reliable conservative criterion shift under low-prevalence visual search conditions, with no consistent effect on sensitivity (d’). Critically, meta-regression revealed that trial-level feedback significantly moderated the magnitude of these shifts; the presence of immediate feedback amplified the LPE, whereas its absence mitigated it. Additionally, the magnitude of the shift was tied to the numerical difference in prevalence, but remained robust across variations in stimulus type, observer expertise, and time constraints. These findings suggest that trial-level feedback reinforces expectations of target scarcity, amplifying the conservative biases that drive the LPE. This resolves theoretical discrepancies between the LPE in visual search and PICC findings in categorization and highlights a potential “intervention paradox”: While generally intended to improve accuracy, standard trial-level feedback may unintentionally exacerbate misses in rare target detection.</p>