American political polarization is traditionally attributed to non-linguistic factors, yet the specific functional-linguistic and cognitive-pragmatic mechanisms in discourse which construct the societal divide remain largely understudied. This article analyzes polarization through functional linguosynergetics, conceptualizing it as a self-regulating “discursive attractor.” Methodologically, the study relies on a 2.9-million-word corpus of multigenre American political texts (2015–2025), combining quantitative statistics with qualitative critical discourse analysis. The data was processed using SpaCy and UDPipe. The findings reveal that polarization operates systemically across many different linguistic strata. Specific linguistic features, including stigma suffixes, loaded partisan labels, invasion metaphors, and agency-obscuring passive voice, function synergistically to reinforce the “us vs. them” dichotomy. This research demonstrates that language does not merely reflect political division but actively drives the discursive system toward a stable, polarized state. By identifying these reinforcing signals, the study offers a novel framework for monitoring and understanding the linguistic catalysts of partisan antipathy.