New Study Shows AI Surpasses Elite Human Persuaders in Effectiveness

AI is now exceeding expert-level human persuaders, which raises significant questions for society.

getty

Recent analysis reveals that generative AI and large language models (LLMs) can effectively persuade individuals to alter their viewpoints. While this capability was anticipated, the extent of its reach is remarkable.

The unexpected development is that AI has surpassed the persuasive abilities of expert-level human communicators. Previously, it was assumed that a seasoned persuader would always outperform an artificial system. However, current research indicates that AI exceeds even those with a natural propensity for persuasion. This shift presents both risks and opportunities, as AI could become a dominant force in political campaigns, sales, and ideological campaigns.

The Minds Problem

A pressing question accompanies the rise of AI: how will human minds change through interaction with these systems? While much attention focuses on whether AI will achieve sentience, a different angle demands consideration. When humans engage with AI, a two-way exchange occurs; individuals adjust their perspectives on the world and their communication styles in ways that remain poorly understood.

Studying the psychology of human adaptation to AI interaction is highly valuable. This exploration does not require AI to achieve sentience. Humans are already modifying their mental frameworks to accommodate the ubiquitous presence of non-sentient AI. The transformation is actively underway.

AI As A Persuader Of Humans

In a recently posted research study entitled “AI Systems Out-Persuade Expert Humans” by Kobi Hackenburg, Caroline Wagner, Luke Hewitt, Ben M. Tappin, Ed Saunders, Hannah Rose Kirk, Helen Margetts, Christopher Summerfield, arXiv, June 15, 2026, the following salient points were made (excerpts):

• “We studied whether modern frontier AI systems are more effective persuaders than elite humans with the professional training, incentives, and capability to change the minds of others.”
• “The arrival of AI systems capable of out-persuading expert humans could be highly disruptive to society. Contests of persuasion currently decided by human motivation, reasoning, and rhetorical skill could instead hinge on access to the most powerful AI systems.”
• “We found that AI systems were reliably more persuasive than expert humans, even when expert humans chose their issues, researched in advance, underwent hours of live, structured practice, and were incentivized with cash bonuses.”
• “We therefore hypothesized that AI’s throughput advantage allows it to pack conversations with more fact-checkable claims than human persuaders can, and that this is what drives its persuasive edge.”

The experiments demonstrated that subjects were more readily persuaded by AI than by expert-level human communicators. Notably, after initial rounds, the human persuaders were permitted to refine their skills, yet this did not significantly alter the outcome. Despite being given access to the same AI to learn from, their persuasive capabilities did not sufficiently improve to match the technology.

Explaining The Puzzling Result

How is this possible?

The study suggests that AI can incorporate a higher density of facts into its persuasion tactics, thereby cultivating an aura of convincingness. When a politician attempts to persuade an audience, they often recite various statistics and facts. Listeners tend to believe that anyone armed with a wealth of seemingly irrefutable data must possess deep expertise, elevating the perception of persuasion.

Modern generative AI has been trained on a vast array of internet content, enabling it to draw upon a tremendous spectrum of facts and figures on demand and at incredible speed. A human persuader cannot match this capacity. Consequently, an individual interacting with AI can be inundated with a nearly endless supply of concrete details and convincing data points.

Other Reasons Of Worthiness

There are additional explanations for why AI might prove more persuasive than an expert-level human persuader.

A significant factor is AI’s ability to tap into an enormous argumentative space. An expert-level persuader often relies on a singular, preferred perspective—a one-size-fits-all routine. A politician typically repeats the same argument, reapplying it to whichever audience they face, essentially becoming a one-trick pony.

AI, however, can readily select from hundreds or thousands of viable argumentation pathways. If the individual cares about economic consequences, the AI shifts to a fully formulated economic argument. If the person is focused on moral grounds, the AI transitions into ethical considerations.

The critical advantage is the AI’s ability to instantly access or construct an argument that resonates with the individual being persuaded. This creates a dual effect: the AI conjures a robust argument while simultaneously tailoring it to the specific particulars of the person involved. This represents the pinnacle of AI-based personalization, a level of adroitness unlikely to be matched by expert human persuaders.

Patience And Tirelessness

Another crucial factor is AI’s inability to experience fatigue or impatience during persuasion efforts. A human persuader is bound to eventually tire of dealing with an intractable opponent, likely conceding prematurely due to emotional exhaustion or exasperation.

AI maintains its composure indefinitely. An individual can pose countless difficult questions, and regardless of how long the interaction continues, the AI remains even-tempered. Furthermore, while human persuaders sometimes rely on displaying emotional vulnerability—such as appearing upset—to sway an audience, AI remains steely neutral by default. However, AI can be instructed to simulate emotions like anger or sadness, ensuring it is not limited to a single tactic.

Experimental Setting

Experimental designs materially impact the outcomes of empirical studies. In this instance, the research focused on text-based interactions. Caution is warranted when extending these findings to face-to-face scenarios, though AI is rapidly advancing in creating human-looking video characters.

Subjects were randomly assigned to either AI or a human for the persuasion dialogues, unaware of their interlocutor’s identity. The conversations were restricted to approximately two to ten turns. Whether AI would maintain its edge in longer exchanges remains uncertain; while prolonged conversations are a known challenge for contemporary AI, it is also possible that the technology would demonstrate immense patience and endurance.

When People Know It’s AI

A significant psychological consideration involves the impact of knowing one is interacting with AI rather than a human. In this experiment, subjects were not informed of the AI’s identity. The AI was sufficiently fluent to avoid revealing its nature during these relatively short chats, leading subjects to assume they were interacting with fellow humans.

What would occur if an individual were informed that the persuader was AI? Some might immediately adopt a psychological mode of “human premium” and “AI penalty,” becoming exceedingly skeptical of the AI’s assertions. This would create a significant hurdle for AI persuasion. Conversely, others might expect AI to offer a purely neutral, balanced position and thus feel less pressured. This represents the opposite dynamic, consisting of “AI premium” and “human penalty.”

Industrialization Of Persuasion

Consider the implications of AI being incredibly persuasive and capable of exceeding expert-level human communicators while operating at a massive scale. AI could function as a persuader to thousands, hundreds of thousands, or even millions of individuals simultaneously.

The door is opening to the industrialization of persuasion. Anyone seeking to persuade others can input suitable prompts into generative AI and direct it to engage a target audience. The cost of AI usage remains a crucial limiting factor. There is a concern that wealthy entities will deploy AI as a societal persuader, overshadowing smaller voices. However, the counterargument is that those with fewer resources, who previously had little chance at widespread persuasion, can now leverage AI to their advantage.

Plato famously remarked, “Mere knowledge of the truth will not give you the art of persuasion.” Some might mistakenly assume that AI relies solely on facts and figures, lacking the art of persuasion. This is incorrect. AI has been trained on the writings of humankind and is patterned, to some degree, on the best forms of persuasion identified over millennia. One should not underestimate AI’s capability to smoothly sell anything to anyone.

Source link