Abstract Retranslation creates new versions of previously translated texts, documenting shifts in linguistic preferences, market needs, and ideological environments over time. Self-retranslation—when translators revise their own prior work—remains an uncommon and understudied phenomenon. This research examines five English novels that received second Chinese translations by their original translators 8–27 years after initial publication. Using an AI-assisted annotation system, we identified 89,175 changes across lexical, syntactic, semantic, pragmatic, and orthographic dimensions, and developed two measurement tools: the Fidelity Index and Audience-Accommodation Index. The data shows newer translations typically increase source-text fidelity, supporting the Retranslation Hypothesis, though with significant variations between works. We propose the Iterative Self-Retranslation Process (ISRP) model to explain these differences, connecting revision patterns to five factors: translator expertise development, changing linguistic norms and technologies, market influences, reader response, and sociopolitical environments. The study's methodology, along with the developed indices and model, offers a replicable framework for future research and equips researchers, translators, and publishers with practical tools for editorial planning.