The Confidence Authenticity Paradox in AI-Assisted English: Evidence from Indonesian EFL Learners

Generative AI has become increasingly embedded in second language use in Indonesia, yet its impact on learners’ communicative experience remains underexplored. Previous research has focused on performance gains and affective benefits, but little is known about how AI-assisted language production shapes a learner’s sense of authenticity. This study examined whether Self-generated and AI-assisted English production differ in their effects on perceived communicative confidence and authenticity among Indonesian EFL learners. A within-subjects survey design was employed with 271 undergraduate students, which evaluated AI-assisted and self-generates English use across learners’ typical English communicative practices. Paired-samples analysis found no significant difference in perceived confidence between AI-assisted and self-generated production. however, perceived authenticity was significantly lower in AI-assisted production, indicating a divergence between identity expression and functional performance. Additionally, correlation analysis revealed confidence and authenticity remained positively related across both conditions. The findings suggest that AI supports communicative performance, but simultaneously disrupts learners’ sense of authorship. This study contributes to human-machine communication by conceptualizing AI as a co-authoring agent that redistributes communicative agency, and highlights the need for pedagogical approaches that fosters identity-safe AI use and spaces in second language environments