<p>The reliable prediction of scholarly recognition of authors remains a persistent challenge in scientometrics, where traditional bibliometric indicators such as citation counts and the h-index provide only partial views of academic achievement. In this study, scholarly recognition is operationalized as the ability to distinguish awardee authors from non-awardees, reflecting formal recognition of scholarly influence. While social network analysis (SNA) introduces structural perspectives through centrality measures, these have often been applied descriptively rather than embedded within predictive frameworks. At the same time, machine learning (ML), deep learning (DL), and large language model (LLM) embeddings have advanced author evaluation, yet typically overlook the relational structures inherent in co-authorship networks. This paper presents a hybrid, centrality-driven ensemble framework for predicting scholarly recognition, integrating normalized network centrality measures with generative AI-based embeddings and calibrated predictive modeling. Using awardee versus non-awardee classification as a test case, we demonstrate that the proposed approach produces highly discriminative author recognition scores on a normalized 0–100 scale, achieving perfect separability between the two groups. The framework consistently outperforms single modality baselines and introduces a novel, interpretable recognition index that unifies structural, semantic, and probabilistic signals. By moving beyond descriptive metrics and siloed prediction models, this work provides a scalable and transparent pathway for author evaluation, offering both theoretical insights and practical tools to support fairer and more context-sensitive research assessment in the era of generative AI.</p>