This Capstone examines how generative AI reshapes the act of writing by treating co-writing itself as a site of inquiry. Using an autoethnographic method, I document my collaboration with ChatGPT across months of drafting, tracing how the model’s predictions influence tone, syntax, and rhetorical choice. The project argues that large language models are not neutral tools but participants in a shared writing process shaped by the linguistic hierarchies embedded in their training data. Through examples of prompting, revision, and negotiation, I show how AI leans toward standardized English, simplifies sentence structure, and reproduces dominant linguistic norms, often in ways that feel fluent but unexamined. Drawing on scholarship from linguistics, rhetoric, AI ethics, and corpus studies, I situate these patterns within broader cultural and technological shifts that are redefining authorship and literacy. My findings suggest that the future of writing will be a form of co-authorship and that ethical engagement requires vigilance, not avoidance: understanding how the model learns, questioning its outputs, and preserving human intention as the final authority. This project therefore offers both a critique of AI’s biases and a practical framework for collaborating with language models responsibly as they become embedded in the work of writing itself.