The integration of AI-powered writing assistants like Grammarly into academic life has fundamentally altered the writing landscape. While these tools provide immediate feedback, their influence on the psychological and procedural aspects of student writing is not well understood. This paper introduces the “Grammarly Effect,” examining it through a sequential explanatory mixed-methods study. The research investigated how real-time AI suggestions affect the writing anxiety and revision habits of undergraduate students. In the quantitative phase (N = 124), a quasi-experimental design using pre- and post-task surveys found a significant decrease in anxiety associated with surface-level correctness (e.g., grammar, spelling). However, the qualitative phase, involving think-aloud protocols and interviews with a subset of students (n = 15), revealed a more nuanced situation. While anxiety over superficial errors diminished, a reliance on the tool fostered a different anxiety related to authorial originality and self-trust. Analysis of revision behaviors showed a consistent pattern of uncritical acceptance of AI suggestions, prioritizing surface-level fixes over deep, rhetorical engagement. The study concludes that the “Grammarly Effect” is a paradoxical phenomenon, highlighting a critical need for pedagogical approaches that foster AI literacy, empowering students to use these tools as aids rather than oracles.