Background: Unexplained symptoms are associated with depression and anxiety. This association is largely based on cross-sectional research of symptoms experienced by patients but not of symptoms presented to the GP. Aim: To investigate whether unexplained symptoms as presented to the GP predict mental disorders. Design and setting: Cross-sectional and longitudinal analysis of data from a practice-based research network of GPs, the Transition Project, in the Netherlands. Method: All data about contacts between patients (n = 16 000) and GPs (n = 10) from 1997 to 2008 were used. The relation between unexplained symptoms episodes and depression and anxiety was calculated and compared with the relation between somatic symptoms episodes and depression and anxiety. The predictive value of unexplained symptoms episodes for depression and anxiety was determined. Results: All somatoform symptom episodes and most somatic symptom episodes are significantly associated with depression and anxiety. Presenting two or more symptoms episodes gives a five-fold increase of the risk of anxiety or depression. The positive predictive value of all symptom episodes for anxiety and depression was very limited. There was little difference between somatoform and somatic symptom episodes with respect to the prediction of anxiety or depression. Conclusion: Somatoform symptom episodes have a statistically significant relation with anxiety and depression. The same was true for somatic symptom episodes. Despite the significant odds ratios, the predictive value of symptom episodes for anxiety and depression is low. Consequently, screening for these mental health problems in patients presenting unexplained symptom episodes is not justified in primary care.