Research on writing development in content and language integrated learning (CLIL) settings is still inconclusive. This study intends to make a contribution in this respect by presenting longitudinal findings on the development of written complexity, accuracy and fluency (CAF) in secondary education CLIL and non-CLIL learners. Participants in the study (N?=?50) are a group of CLIL students (N?=?30) learning science or social science through the medium of English and a group of comparable non-CLIL learners (N?=?20) of English. Analytical CAF measures are used to investigate their performance in a written composition at four data collection times over three years. Within-group results show significant differences in the scores obtained in most of the CAF measures for CLIL learners at the different data collection times, compared to significant differences only in lexical complexity and accuracy for non-CLIL participants. Results of between-group comparisons controlling the hours of exposure are mixed: non-CLIL students progress significantly more in lexical complexity than their CLIL peers, although the latter tend to obtain higher scores overall. In sum, evidence suggesting the greater effectiveness of the CLIL approach is found.