This study inquired into the structure and predictors of out-of-school learning assistance among adolescents. It used longitudinal data gathered from up to 4936 Polish lower-secondary school students. The analysis of the assistance structure, performed with latent class analysis, showed that about 50% of students belonged to the latent class with no assistance. All other students belonged to one of four latent classes, representing types of assistance. These latent classes were: frequent non-paid assistance in all subjects, moderate non-paid assistance in all subjects, non-paid assistance in selected subjects, paid and non-paid assistance in mathematics. The analysis of predictors of assistance provision showed that 12 out of 14 hypothesised factors affected membership probability (MP) in at least one latent class. A lower level of past school achievement raised MP in all the classes. Students' gender, intelligence, intellectual helplessness, learning disability, educational aspirations, family affluence, level of parental education, and school location predicted MP in selected classes.