AI-driven recommendation algorithms increasingly shape user experience on social media, raising concerns about transparency, accountability, and user agency. This paper presents a comparative analysis of Explainable AI (XAI) implementations on Facebook, Instagram, TikTok, and Twitter/X. Using a structured framework, we assess two key dimensions: Explanation Adequacy, defined by clarity, specificity, relevance, and verifiability, and Explanation Actionability, defined by proximity, granularity, and reversibility of controls. Our evaluation combines feature audits, cross-platform comparisons, and rubric-based scoring. Results show Facebook provides the strongest balance of adequacy (4/5) and actionability (4/5), Twitter/X offers limited adequacy (2/5) but moderate actionability (3/5), TikTok demonstrates strong actionability (4/5) but generic explanations, and Instagram performs moderately ($3 / 5$on both dimensions). We further extend the analysis by linking adequacy and actionability to perceived usefulness, trust, satisfaction, and algorithmic scepticism. Findings highlight tensions between algorithmic sophistication, user comprehension, and engagement optimization, while also revealing regulatory implications under GDPR and the DSA. This work contributes a standardized evaluation framework for XAI in social computing, empirical evidence of platform disparities, and practical design guidelines for enhancing transparency and user empowerment in recommender systems.