This research explores the issue of synonymy in Arabic through the lens of corpus-based pragmatics, examining particularly the boundaries of semantic identity of near-synonymous lexical items. The qualitative descriptive approach is based on a mixed corpus of formal news texts and X platform posts, allowing the study of synchronic natural language use in a variety of contexts. Five sets of near-synonyms - (قال/ذكر qāla/dhakara "say/mention"), (رأى/نظر raʾā/naẓara "see/look"), (قتل/اغتال qatala/ightāla "die/assassinate"), (بدأ/شرع badaʾa/sharaʿa "begin/start"), and (كبير/ضخم kabīr/ḍakhm "big/huge") - were analysed from four perspectives: semantic, pragmatic, collocational, and equivalence. It emerges that in Arabic, synonymy does not translate into semantic equivalence, but rather into a contextualized and restricted approximation based on collocational, pragmatic, and textual factors. Words have their own usage environment that controls their distribution and the extent to which they can be swapped. The research also shows that semantic equivalence is not a lexical attribute, but a relative, contextual, and emergent process that occurs only when several semantic and pragmatic criteria are fulfilled. This study, by reconceptualizing synonymy in a usage-based (pragmatic) perspective, shows the role of corpus data analysis in discovering subtle meanings. This paper adds to the understanding of meaning in Arabic and has implications for translation and discourse analysis.