Online interactions and novel digital innovations have profoundly transformed how individuals connect in virtual spaces. Although dependence on the web for routine activities, leisure and social engagements continues to grow among enterprises, organizations and users, digital social communication still poses persistent obstacles. The improper utilization of networking platforms can facilitate the dissemination of false information. This phenomenon has propelled the issue of false news into the spotlight of public discourse. The immense volume of user-generated data and the rapidly shifting linguistic patterns prevalent on social media channels render conventional techniques largely inappropriate. This study introduces a data-centric methodology that harnesses practical artificial intelligence to identify and counteract false news, employing sophisticated natural language processing (NLP) strategies like Word2Vec. It incorporates an enhanced implementation of the crayfish optimization algorithm (PSO) for fine-tuning XGBoost classifier’s parameters, yielding encouraging outcomes in terms of achieved classification accuracy. A side-by-side assessment involving multiple modern optimizers on a genuine dataset confirms the strength of the proposed framework, reaching a peak accuracy rate of 95.22% with the highest achieving XGBoost models.