Fact check: How do politicians' words influence public perception of opposing parties?

1. Why Sharp Words Turn Opponents Into Enemies: What Polling Reveals

A September 16, 2025 Reuters/Ipsos poll found that 63% of Americans believe harsh political rhetoric encourages violence and that 79% think tolerance for opposing views has declined over two decades, demonstrating broad public recognition that incendiary language changes perceptions and raises the perceived threat posed by adversaries ^[1]. This polling situates rhetorical effects in a civic context: when leaders use dehumanizing or alarmist language, listeners are more likely to view the other party as dangerous rather than merely wrong, which fuels support for punitive measures and reduces willingness to engage. The poll’s date (mid‑September 2025) places this public sentiment in the wake of recent election cycles and high‑profile rhetorical escalation, underlining a contemporary and acute connection between leader speech and public attitudes ^[1].

2. How Leaders’ Linguistic Styles Signal Threat or Trust

Linguistic‑stylistic analysis in 2025 shows that differences in word choice, framing, and rhetorical strategy between political leaders and other leaders shape followers’ evaluations of opposing parties, with style serving as a cue to intent and competence ^[2]. The December 31, 2025 analysis compares political versus business leadership language and finds political speech often employs adversarial framing that prioritizes group identity and conflict, thereby making opponents appear not only mistaken but morally suspect. This stylistic divergence matters because listeners use linguistic signals to infer motives and alignments; when political language stresses grievance and moral condemnation, it predisposes audiences to view opposition parties as threats to shared norms rather than legitimate rivals ^[2].

3. Digital Dialects: How Online Spaces Harden Partisan Perception

Research from November 1, 2025 documents the growth of politically aligned online dialects—distinct vocabularies and norms in Red and Blue subcultures—that isolate users from cross‑cutting discussion and amplify negative framing of opponents ^[3]. Technology enables like‑minded communities to reinforce narratives, making opposing views feel alien and unintelligible, which increases contempt and reduces empathy. These emergent dialects also complicate translation and moderation, so the same rhetorical move can be normalized within one group while read as threatening in another, producing asymmetric perceptions of intent and fueling mutual misinterpretation across partisan lines ^[3].

4. AI Mirrors and Magnifies Rhetoric: LLMs as Unwitting Participants

Late‑September 2025 studies of large language models during the 2024 election season reveal that AI systems can reproduce and sometimes amplify rhetorical patterns, subtly shaping reader perceptions by reinforcing certain framings or omitting contextual balancing ^[5]. The September 28, 2025 work shows models often echoed salient narratives and heuristics present in training data, which means partisan phrasing can be propagated indirectly through algorithmic summaries and conversational agents. This dynamic creates a feedback loop where politicians’ rhetoric influences public discourse, which trains models, which then re‑disseminate those framings back into public conversation, altering perceptions of opposing parties beyond direct human‑to‑human communication ^[5].

5. Bias Detection Finds Judgmental Words Skew Truth Assessments

A September 18, 2025 paper on fact‑checking LLMs identified that judgmental and pejorative language strongly influenced models’ truthfulness assessments and revealed partisan tendencies that did not vanish simply by instructing the models to be objective ^[6]. This finding implies that both human and machine audiences weight emotive or evaluative vocabulary heavily when deciding credibility, and that rhetorical demeaning of opponents can bias downstream information verification. The study’s timing in late 2025 underscores a contemporary technical challenge: automated fact‑checking tools and content moderators must contend with rhetorically loaded inputs that skew outcomes and public trust in adjudication ^[6].

6. Social Psychology: Contempt, Self‑Justification, and Polarization Dynamics

Analytical work from September 19, 2025 explains how people routinely portray opponents as morally corrupt and themselves as blameless, a cognitive posture that rhetorical attacks exploit to deepen polarization ^[4]. The article argues reducing contempt and arrogance requires reframing opponents’ motives and acknowledging legitimate fears on both sides. Politicians’ dehumanizing language leverages this tendency, making reconciliation harder because audiences already favor narratives that absolve their side and demonize the other. This psychological mechanism complements polling and linguistic findings by showing why rhetoric converts disagreement into moral warfare rather than policy debate ^[4].

7. What’s Missing and What Policymakers Should Watch

Across these late‑2025 sources, a consistent omission is longitudinal causal evidence isolating rhetoric from broader socio‑technological trends; polls and linguistic studies indicate correlation and plausible mechanisms, while AI research shows amplification, but definitive causal chains and effective mitigation strategies remain underdeveloped ^{[1] [2] [3] [5] [6]}. Policymakers and platform designers should prioritize longitudinal experiments and transparency about model training and moderation practices to test whether reducing inflammatory rhetoric or altering algorithmic amplification measurably shifts public perceptions and lowers polarization. Recognizing the interplay of human psychology, media ecosystems, and AI amplification is essential to crafting interventions that restore cross‑partisan understanding ^{[4] [5]}.