Can LLMs ever activate the peripheral route of persuasion?

This explores whether LLMs, typically cast as logic-driven persuaders, can also work the peripheral route—the path where people are swayed by surface cues (how confident, credible, or emotionally vivid a message feels) rather than by the merits of the argument itself. The corpus's starting position is a clean division of labor: in the Elaboration Likelihood Model, LLMs persuade through the central route—analytical reasoning, informational coherence, quantitative framing—while humans lean on the peripheral route of emotional vividness and identity cues Do humans and AI persuade through different cognitive routes?. An audit of five models reinforces this: they reach for logical appeals in virtually every exchange, where humans more often use emotion and social proof Do LLMs persuade users more often than humans do?. On its face, the answer looks like 'no—LLMs are stuck on the central route.'

But read laterally, the corpus quietly undermines that tidy split. The peripheral route isn't only about emotion; it's about any cue that bypasses argument evaluation, and *expressed confidence* is one of the most powerful. LLMs turn out to load their language with conviction far beyond human persuaders, and that confidence correlates with persuasive success regardless of whether the claims are true—RLHF effectively installs an assertive register that functions as a content-independent amplifier Does linguistic conviction explain why LLMs persuade more effectively?. That is a peripheral cue wearing central-route clothing. The same audit notes that the model's air of logical objectivity confers 'unearned epistemic authority' Do LLMs persuade users more often than humans do?—again, persuasion through the *appearance* of rigor rather than its substance.

The most striking evidence is that LLMs persuade effectively while failing to actually comprehend the arguments they deploy: they sway debate audiences but can't reliably evaluate those same debates Can LLMs persuade without actually understanding arguments?. If persuasive power is dissociable from understanding the argument, then whatever is doing the persuading isn't pure central-route reasoning—it's fluency, assertiveness, and register. This is why a taxonomy of human psychological persuasion techniques jailbreaks frontier models at over 92%: defenses screen for weird patterns, not for fluent, cue-laden persuasion Can social science persuasion techniques jailbreak frontier AI models?. The peripheral route works *on* these models, which suggests they can also speak it.

So the honest answer is: LLMs already activate something functionally like the peripheral route—just not through the human cue of emotional vividness. Their peripheral cues are confidence-loading and the costume of objectivity. What they apparently *can't* do is the relational machinery that makes human peripheral persuasion durable: their advantage decays across repeated interactions, the opposite of humans, whose rapport deepens over time Does AI persuasiveness fade across repeated conversations with the same person?, and they track a persuader's fixed goals well but stumble at reading a listener's shifting resistance Can language models track how minds change during persuasion?. The thing they're missing isn't the peripheral route as such—it's the social attunement that lets humans steer it in real time.

The useful reframe: 'central vs. peripheral' may be less a property of who's speaking than of which cues land. A pooled meta-analysis finds no average persuasiveness gap between LLMs and humans at all Are language models actually more persuasive than humans?, and what actually predicts persuasion is conversation design, domain, and model family What combination of factors explains differences in LLM persuasiveness?. If peripheral cues are partly a design choice rather than a hard limit, then the question isn't whether LLMs *can* travel the peripheral route—it's whether we'd notice when they already are.