Background: Generative artificial intelligence has transformed literary production across publishing, education, cultural industries, and digital platforms while intensifying disputes over creativity, attribution, ownership, accountability, and platform power. Objective: This study examines how creative agency, constraint regimes, and authorship governance are configured across contemporary machine co-authorship discourse and what these configurations reveal about institutional legitimacy. Method: Using qualitative comparative documentary analysis, this study coded forty publicly accessible records comprising thirteen primary documents, seventeen secondary sources, and ten contextual materials published or current between 2021 and 2026 through explicit decision rules and evidence notes. Results: Creative agency was most frequently represented as an integrated human–machine–platform chain combining initiation, prompting, generation, revision, and mediation. Institutional constraints dominated the corpus, while aesthetic, technical, legal, economic, and data-related conditions operated as enabling, limiting, or simultaneously ambivalent forces. Human accountability and platform power were substantially more prevalent than contractual output ownership, indicating that creative participation, legal authorship, and institutional control remain unevenly distributed. Implication: Legitimate machine co-authorship requires transparent contribution, proportionate attribution, human editorial responsibility, accountable training practices, and governance capable of protecting creative diversity. Novelty: This study contributes an integrated framework that connects distributed agency, creativity-under-constraint, and authorship politics within a single document-level analytical architecture.