A recent study of the uses of MT by academics and students (Benites et al, 2021) has shown that they have already become widespread for non-linguists in their every-day professional activities --e.g. writing emails, research presentations, articles, assignments, assessments, etc. Therefore, it becomes necessary for this “lay” public to develop cognitive skills for MT literacy (Bowker & Ciro, 2019), despite the fact that they do not always have enough linguistic knowledge for an informed criticism of MT outputs. Indeed, trust in imperfect MT has been reported and even measured, based on adequacy and fluency (Martindale & Carpuat, 2018). However, existing studies have not taken into account uses in a specialized field, related to professional practice, for the purpose of translating and not just gisting (Nurminen 2021). To consider this issue we have set up a study in a specific communicative situation: non-native academics and students required to use English in the field of biotechnology for health. Dealing with such ethically-sensitive issues as gene-editing or cellular manipulation, this is a scientific field where finding the right word is crucial. Will these non-native speakers using MT for professional purposes have the necessary skills to make the best use of MT output and how can we, as translation specialists, help them to hone those skills? The overarching question we will be trying to answer in this article is the following: to what extent can translators and MT users speak the same language about what MT does wrong? Based on the detailed analysis of a specialized corpus of English MT output on gene-editing (from French source texts, using two free online NMT engines) we designed a short survey to test academics’ and students’ perception of poorly translated segments. From these results, we proposed a classification of errors in our corpus according to how likely it is they should be spotted by users from our study. This perceptibility index (1 perception likely 2 perception rather unlikely 3 perception highly unlikely) would enable highlighting of errors that might remain in published texts, namely in scientific papers by non-native speakers. Through the analysis of errors according to this index and to their impact on communication, we also offer insight as to how language might evolve if machine translation is used without particular caution in this specialized academic field. Bowker, L., & Ciro, J. B. (2019). Machine translation and global research: Towards improved machine translation literacy in the scholarly community. Emerald Group Publishing. Benites, A. D., Kureth, S. C., Lehr, C., & Steele, E. (2021). Machine translation literacy: a panorama of practices at Swiss universities and implications for language teaching. CALL and professionalisation: short papers from EUROCALL 2021, 80. Martindale, M., & Carpuat, M. (2018). Fluency over adequacy: a pilot study in measuring user trust in imperfect MT. In Proceedings of the 13th Conference of the Association for Machine Translation in the Americas (Vol. 1: Research Track, pp. 13-25). Association for Machine Translation in the Americas. Nurminen, M. (2021). Investigating the Influence of Context in the Use and Reception of Raw Machine Translation. PhD Dissertation, Tampere University, Finland. (https://trepo.tuni.fi/handle/10024/135446 consulted November 22, 2022).