Generative artificial intelligence (GenAI) is rapidly transforming postgraduate research communication, creating opportunities for academic assistance while also posing risks to source authenticity and intellectual stewardship. Based on an empirical multi-site meta-synthesis of N=4,280 doctoral and early-career researchers across eight major universities, this study establishes and validates the Accountable Scholarly Augmentation framework. Findings show GenAI effectively supports low-level tasks like outlining and stylistic framing. However, unguided use creates an "Overconfidence Gap," where high prose fluency masks factual errors and citation hallucinations (up to 38.2% unverified claims). Multivariate analysis reveals that punitive policy warnings fail to reduce manuscript defects. Conversely, integrated responsible AI programs significantly decrease pre-submission desk rejections from 41.8% to 12.6% (β=.52,p<.001). The study emphasizes shifting research integrity governance and graduate education from punitive policy enforcement to a developmental publication-readiness framework.