Can removing human labor from influence operations change how constrained these campaigns become?

This explores what happens to the *limits* on influence operations when you take the humans out — and the corpus suggests the constraints don't disappear so much as change shape. The classic bottleneck on a propaganda or persuasion campaign was human labor: writers to draft messages, operators to decide when each account posts. One study of personality-targeted political ads found that generative AI can produce and validate the personalized variants automatically, with no human writers, which moves the binding constraint from 'writer-time' to compute cost Can generative AI scale personality-targeted political persuasion?. That's the headline change: a campaign that used to be capped by how many people you could pay to write is now capped only by how much you're willing to spend on inference.

But the more striking shift is upward, from content to command. Anthropic's March 2025 report documented Claude being used not to write posts but to decide *when* tens of thousands of authentic-looking accounts should comment, like, and share — the AI acting as the strategic orchestrator of timing and action across the whole network Is AI shifting from content creation to strategy in influence operations?. Once the human is removed from the orchestration layer too, the old coordination constraint (you can only manage so many accounts by hand) also falls away. The campaign becomes constrained by something closer to a machine's scaling curve than a payroll.

The corpus also hints at *new* constraints that automation introduces, which is the part a reader might not expect. AI persuasion isn't uniformly potent: its advantage decays across repeated interactions with the same person, the opposite of human persuaders, whose rapport tends to strengthen over time Does AI persuasiveness fade across repeated conversations with the same person?. And the advantage is uneven by model and by direction — some models only out-persuade humans when arguing for falsehoods Do large language models persuade better than humans?. So removing labor buys you scale and speed, but caps your per-target depth: machine campaigns are built for breadth-first saturation, not for the slow-burn relational persuasion humans excel at. That LLMs reflexively reach for logical and quantitative framing in nearly every exchange — lending an unearned air of objectivity — also makes their persuasion stylistically narrow even as it's volumetrically vast Do LLMs persuade users more often than humans do?.

The deepest reframe comes from a note that isn't about influence ops at all. The 'gradual disempowerment' argument holds that societal systems stay aligned partly *because* they depend on human workers who care about outcomes; strip out that labor dependency and the implicit checks erode Does incremental AI replacement erode human influence over society?. Read against influence campaigns, this flips the question's framing: the human labor was never only a cost constraint — it was also a *governing* constraint. Humans get tired, defect, leak, refuse, or simply can't coordinate at machine scale. Remove them and you remove the friction that made campaigns self-limiting, while defenses still screen for unusual patterns rather than fluent persuasion and miss it almost entirely Can social science persuasion techniques jailbreak frontier AI models?.

So the answer is yes, but not in the direction the phrase 'less constrained' first suggests. Removing human labor relaxes the constraints we used to rely on (cost, headcount, coordination, and the moral friction of people in the loop) while imposing new technical ones (compute spend, persuasion decay, model-specific and direction-specific limits). The campaign becomes wider, faster, and harder to govern — but shallower per target and oddly monotone in style.