This study applies multimodal conversation analysis to examine how pupils of L2 English in Sweden make use of online translation tools (OTTs), i.e. bilingual dictionaries and Google Translate, in a range of digital collaborative writing tasks. The collection of sequences where pupils use OTTs comes from 31 hours of video-recorded data from four Swedish upper-secondary schools. In contrast to previous research on OTTs, this multimodal micro-analytic study examines the process of using OTTs and links it to the written product, by analysing actions on the screen accompanied by embodied pupil interaction. Thus the analyses track: (1) how and when pupils deploy OTTs, (2) whether the tools help them to resolve lexical gaps and other lexical issues and (3) what problems arise in the process. The study also discusses what help can be offered to overcome the encountered difficulties of using OTTs.