This research tackles the complex issue of Named Entity Disambiguation (NED) by harnessing the power of Large Language Models (LLMs), such as Generative Pre-trained Transformers (GPTs), and integrating these with structured information from knowledge graphs like Wikidata. The overarching goal is to markedly enhance NED accuracy through a synergistic approach that combines the predictive prowess of advanced LLMs with the detailed semantic repositories of KGs. Our key innovation is the formulation of a novel methodology that employs SPARQL queries to methodically retrieve potential Uniform Resource Identifiers (URIs) from Wikidata. These URIs serve as candidate references for named entities. We then apply cutting-edge generative AI techniques to accurately match each entity’s mention within texts to the most contextually relevant URI. This refined approach allows for a more sophisticated, context-aware resolution of entity references. We conducted extensive experiments to evaluate our method, which indicates a substantial achievement in performance, as evidenced by precision, recall, and F-measure. Impressively, our model achieves an F-measure of 96% on the Wikidata-Disamb dataset, a meticulously curated benchmark specifically developed for robust, scalable evaluation of NED performance leveraging Wikidata entries.