Here, we investigated the association of different emotion regulation (ER) indices with symptom severity across a large transdiagnostic sample of patients with emotional disorders (cross-sectional approach) and the predictive validity these ER indices have for the outcome of routine care CBT (longitudinal approach). We assessed the trait-like use of adaptive (reappraisal) and maladaptive (suppression, externalizing behaviors) ER strategies via questionnaire as well as the situational ability to regulate emotions with an experimental ER paradigm. Psychopathology was assessed dimensionally using the depression, anxiety, and stress scale. Cross-sectionally symptom severity was predicted by less trait-like use of adaptive and more trait-like use of maladaptive ER strategies, but no associations were found for situational ER ability. This association was more pronounced for depression and stress symptoms rather than anxiety symptoms. In a striking dissociation, the longitudinal analyses revealed the reverse picture: Better situational ER ability, but not trait-like use of ER strategies was associated with less symptom severity after the CBT treatment. Our data argues in favor of a distinction between trait-like and situational ER abilities in individuals with emotional disorders, highlighting challenges in applying adaptive ER strategies in daily life despite demonstrating intact ER skills in experimental settings. Our findings also inform transdiagnostic models of psychopathology and suggest that distress/depression rather than anxiety symptomatology to be driving forces for the occurrence of ER deficits across the depression/anxiety disorders spectrum.