With the increased frequency in the use of online translation tools (OTTs), such as Google Translate, in the foreign language classroom, teachers are reported to be concerned about students’ overdependence on these tools, their potential detrimental effects on learning, ethical issues of using them inappropriately and students’ lack of skills to use them critically. Thus, this study sets out to investigate the basis of teachers’ scepticism empirically by analysing OTT search sequences that display critical interactional strategies, where students reject the first translation option(s) offered by an OTT. The data analysed come from 22 h of video data of collaborative writing tasks, tracked using multimodal conversation analysis. The findings revealed five critical OTT strategies deployed to resolve emergent lexical issues. However, the pairs of students varied greatly in their use and range of critical OTT strategies, suggesting the need for explicit training in using different strategies.