Abstract
<jats:title>Abstract</jats:title> <jats:p>When neural machine translation was integrated into Google Translate in 2016, higher‐quality machine translation (MT) became available to serve a range of purposes, including learning a second language (L2). Free online MT systems such as Google Translate and DeepL, while impressive in their capabilities, are nonetheless not infallible. MT performance is notably modulated by genre and by language pair, with closely related languages and high‐resource languages (for which extensive bilingual corpora exist) benefiting from greater accuracy. MT can also reproduce the gender, racial, or ethnic biases and stereotypes present in corpora. Insofar as language learners are aware of these limitations, ideally receiving training to address the issues, MT can be effectively leveraged for language acquisition. Incidental L2 learning is a potential byproduct of MT use for professional or personal communication, for example, in the workplace or for tourism. Intentional learning is more closely associated with formal classes or with autonomous, self‐directed learners. MT can be accessed for L2 compositions, with learners comparing their own and MT formulations to promote noticing and metalinguistic awareness. MT can also be used in activities that target specific L2 features appropriate to learners' age and proficiency. Unfortunately, MT use at times complicates formative and summative assessment, making it hard to identify learner needs and achievements. Nonetheless, the many features of online MT (e.g., dictionary, spellchecker, and text‐to‐speech and speech recognition functionalities) make it an invaluable, multifaceted resource for independent and classroom‐based L2 learning.</jats:p>