‘I want the text to still be me’

Abstract This study foregrounds the potential impact of ChatGPT usage on second language (L2) learners’ voice. L2 learners’ perceptions of ChatGPT’s interference with their self-representation in writing and the relationship of these perceptions with future intentions to use ChatGPT are investigated. Following a mixed-methods approach, questionnaire data from undergraduate students ( N = 162) after two English for academic purposes courses, where they used ChatGPT as L2 academic writing support, were collected and analysed. About half of the participants were ambivalent; the remaining half were divided equally in their perceptions of ChatGPT as (not) a threat. The more threatening ChatGPT was perceived to be, the less inclined participants were to use it again. Ambivalence is influenced by usage, genre and GenAI advancements. Over-reliance, homogenisation and devaluation of human voice relate to the perception of? ChatGPT as a threat. A lack of perceived threat derives from ChatGPT’s mechanical voice, human agency, and different standards for acceptable convergence.