This study aims to describe students' perceptions of the role of Grammarly and QuillBot as artificial intelligence (AI)-based tools for providing feedback in writing instruction. Employing a qualitative descriptive approach, the research was conducted at three senior high schools in Lhoksumawe, Indonesia, with 12 purposively selected respondents based on their prior experience using both applications. Data were collected through semi-structured interviews lasting 30 - 45 minutes, analyzed using thematic analysis, and validated through data triangulation and member checking. The findings indicate that the majority of students perceived Grammarly as effective in detecting grammatical, spelling, and punctuation errors, with its explanatory features aiding their understanding of error patterns. QuillBot was regarded as superior in paraphrasing, enriching linguistic variation, and reducing plagiarism risks. Both tools accelerated the revision process, enhanced confidence, and improved language accuracy. However, identified challenges included limitations of free features, potential meaning distortions in paraphrasing, cultural-context mismatches, and risks of overreliance. The findings underscore the importance of a "human-in-the-loop" model that integrates AI's speed and consistency with teacher validation and contextual assessment. This blended feedback strategy holds potential for improving writing quality while fostering students' critical thinking and self-evaluation skills. The study recommends AI literacy training, restricted use in final revision stages, and classroom discussions based on AI-generated corrections to promote more reflective and effective writing instruction.