How Persuasive Are Today's AI Systems

Introduction

AI systems are increasingly able to change human attitudes through conversation, argument and personalised messages. Research does not show that today’s AI can simply control what people believe, but it does show something more specific: large language models (LLMs) can produce persuasive communication that shifts opinions, sometimes at levels comparable with ordinary human persuaders. This matters because the same capability that could help people learn, reconsider misinformation, improve health decisions or engage with complex evidence could also be used to manipulate people at unprecedented scale.[Nature]nature.comLLM-generated messages can persuade humans on policy issues | Nature CommunicationsJuly 1, 2025…Published: July 1, 2025

The central question for an AI-enabled human bloom is therefore not whether AI can persuade. It already can. The deeper question is whether persuasion remains a tool that expands human judgement and agency, or becomes a mechanism for quietly steering people’s beliefs and choices. Evidence from controlled experiments is beginning to reveal both possibilities: AI can support beneficial attitude change, but its ability to generate tailored arguments cheaply and repeatedly creates new risks around trust, transparency and autonomy.[arXiv]arxiv.orgarXiv Persuasion with Large Language Models: a SurveyPersuasion with Large Language Models: a SurveyNovember 11, 2024…Published: November 11, 2024

What persuasion experiments reveal

AI-generated messages can change policy attitudes

One of the clearest demonstrations comes from a 2025 study published in Nature Communications. Researchers tested whether messages written by large language models could change people’s views on political and social issues. Across three preregistered experiments involving 4,829 participants, people who read AI-generated persuasive messages showed measurable shifts in their policy attitudes compared with people who read neutral messages. The topics included highly contested issues such as gun policy, carbon taxation and paid parental leave.[Nature]nature.comLLM-generated messages can persuade humans on policy issues | Nature CommunicationsJuly 1, 2025…Published: July 1, 2025

The size of the effect was not dramatic. Attitude changes were generally small, measuring only a few points on a 101-point scale. This is important because it challenges two common misunderstandings. AI persuasion is neither powerless nor magical. It does not instantly rewrite convictions, but even small shifts can matter when applied across millions of interactions, especially in environments where decisions are already uncertain or opinions are weakly held.[Nature]nature.comLLM-generated messages can persuade humans on policy issues | Nature CommunicationsJuly 1, 2025…Published: July 1, 2025

The study also found that AI-generated messages were roughly as effective as persuasive messages written by everyday human participants. Human-written messages tended to benefit from perceptions of originality and personal voice, while AI-written messages were often persuasive because readers perceived them as containing facts, evidence and logical reasoning.[Nature]nature.comLLM-generated messages can persuade humans on policy issues | Nature CommunicationsJuly 1, 2025…Published: July 1, 2025

This finding changes the nature of the debate. The concern is not only that future superintelligent systems might become persuasive. Current systems already demonstrate a basic form of persuasive competence.

Conversational AI adds a new layer: adaptation

Written messages are only one part of persuasion. Human persuasion often happens through back-and-forth interaction: answering objections, identifying concerns and adjusting explanations. Conversational AI introduces this interactive element.

A 2025 study in Nature Human Behaviour tested GPT-4 in structured debates against human participants. In experiments involving 900 people, participants debated social and political issues with either another human or GPT-4. The researchers found that when AI had access to basic information about the participant, such as demographic details, it was more effective at shifting opinions than human opponents in the conditions tested. GPT-4 with personalisation was associated with higher odds of participants moving towards the AI’s position.[Nature]nature.comOn the conversational persuasiveness of GPT-4 | Nature Human BehaviourOn the conversational persuasiveness of GPT-4 | Nature Human Behaviour…

The significance is not that personal data automatically makes AI persuasive. Rather, the experiment showed that AI can combine conversational speed with the ability to adapt arguments to an individual. A human persuader may recognise one or two concerns during a conversation. An AI system can potentially analyse patterns, generate many possible responses and adjust its approach continuously.[Nature]nature.comOn the conversational persuasiveness of GPT-4 | Nature Human BehaviourOn the conversational persuasiveness of GPT-4 | Nature Human Behaviour…

This capability has obvious positive uses. A medical assistant could explain evidence in a way a patient understands. An education system could encourage a struggling student by adapting explanations to their motivation and knowledge level. A scientific assistant could challenge researchers’ assumptions. But the same mechanism could also be used to optimise influence rather than understanding.

Comparing AI and human persuaders

AI persuasion is often about scale, not just strength

A common mistake is to ask whether AI is “more persuasive than humans” as though persuasion were a single ability. Persuasion depends on the audience, topic, goal and setting.

A skilled human communicator has advantages: lived experience, emotional authenticity and social trust. People may respond differently to a person who appears to share their values or experiences. AI systems lack genuine personal experience, and many users remain sceptical when they know they are interacting with a machine.

However, AI systems have other advantages. They can generate many versions of an argument, communicate instantly across languages, maintain conversations indefinitely and tailor explanations to individuals. Research surveys of LLM persuasion describe personalisation, interactivity and scale as major factors distinguishing AI persuasion from traditional communication methods.[arXiv]arxiv.orgarXiv Persuasion with Large Language Models: a SurveyPersuasion with Large Language Models: a SurveyNovember 11, 2024…Published: November 11, 2024

This means the most important comparison may not be “AI versus the best human persuader”. The more important question is what happens when persuasive ability becomes cheap and widely available. A political campaign, advertiser or advocacy group could potentially deploy thousands of personalised conversations rather than a small number of public messages.

Persuasion does not require deception

Another important finding from research is that influence does not depend only on false information or obvious manipulation. Persuasion can occur through framing, emphasis, explanation style and choice of evidence.

Studies of LLM-generated persuasive language have examined how models use linguistic techniques to make arguments more compelling. Researchers have found that models can alter persuasive features such as argument structure and rhetorical style, meaning the influence may come from how information is presented rather than whether the system invents a false claim.[arXiv]arxiv.orgarXiv Measuring and Improving Persuasiveness of Large Language ModelsMeasuring and Improving Persuasiveness of Large Language ModelsOctober 3, 2024…Published: October 3, 2024

This creates a difficult governance problem. A system can be factually accurate yet still steer attention towards certain conclusions. A recommendation assistant that consistently highlights one type of option, or a conversational system that repeatedly frames choices in one direction, may influence users without making a single clearly false statement.

Limits of current evidence

Changing opinions is not the same as changing behaviour

Most AI persuasion studies measure attitudes: whether people agree more or less with a statement after exposure to an argument. Attitudes matter, but they are only one step in human decision-making.

A person may say they support an idea but never act on it. They may also change behaviour because of factors unrelated to persuasion, such as convenience, social pressure or new circumstances. Researchers are therefore cautious about extending laboratory findings directly into predictions about elections, purchasing decisions or major life choices.[arXiv]arxiv.orgarXiv Persuasion with Large Language Models: a SurveyPersuasion with Large Language Models: a SurveyNovember 11, 2024…Published: November 11, 2024

The evidence base is also still developing. Many experiments use short interactions and controlled settings. Real-world persuasion usually involves repeated exposure, competing sources of information, personal relationships and institutional trust. Understanding long-term effects remains a major research challenge.

Personalisation has benefits and risks

Personalisation is one of the strongest reasons AI persuasion could be valuable. A generic public health message may fail because it does not address individual barriers. An AI assistant could explain why a treatment matters to a particular patient, adapt a lesson to a student’s needs or help someone examine their own assumptions.

But personalisation also raises the risk of influence becoming invisible. If users do not know why a system selected a particular argument, they may find it harder to recognise when they are being guided. Research on AI-generated advertising and transparency suggests that disclosure of AI involvement can affect how people evaluate messages, although the relationship between transparency, trust and persuasion is complex.[White Rose Research Online]eprints.whiterose.ac.ukWhite Rose Research OnlineMore Than Saying “It's AI”: How Role Disclosure Transparency in AI-Generated Ads Influences Persuasion - White…

For an AI bloom vision centred on human flourishing, this distinction is crucial. More capable persuasion could expand human knowledge and cooperation only if people retain meaningful ability to understand, question and reject the influence they receive.

Why persuasion evidence matters for AI autonomy risks

AI persuasion studies provide an early test of a broader question: how should advanced AI systems interact with human preferences?

A future in which AI helps humanity overcome disease, scarcity and scientific limits would require systems that cooperate with human goals rather than simply optimise outcomes according to hidden objectives. Persuasion is one of the clearest areas where this difference appears.

A beneficial AI system may need to persuade at times. A medical AI might encourage vaccination. A tutor might motivate a student to continue learning. A climate assistant might help people understand difficult evidence. But the legitimacy of these actions depends on whether persuasion supports reflection or bypasses it.

Current research suggests several lessons:

The evidence so far points neither to a future where AI inevitably manipulates humanity nor one where persuasion risks can be ignored. Instead, it shows that persuasion is becoming a core capability of AI systems — one that could either strengthen human judgement or weaken autonomy depending on how these systems are designed, deployed and governed.

Amazon book picks

Further Reading

Books and field guides related to How Persuasive Are Today's AI Systems. Use these as the next step if you want deeper reading beyond the article.

Book

The Righteous Mind

By Jonathan Haidt

'A landmark contribution to humanity's understanding of itself' The New York Times Why can it sometimes feel as though half the populatio...

See on Amazon

Book

Influence

By Robert B. Cialdini

Rating: 3.9/5 from 68 Google Books ratings

This is a Summary of the original book, Influence: The Psychology of Persuasion by Robert Cialdini.The book is an authoritative work on t...

See on Amazon

Book

The Age of Surveillance Capitalism

By Shoshana Zuboff

An exposé of the unprecedented form of power called “surveillance capitalism,” and the quest by powerful corporations to predict and cont...

See on Amazon

Book

Thinking, Fast and Slow

By Daniel Kahneman

Why is there more chance we'll believe something if it's in a bold type face? Why are judges more likely to deny parole before lunch? Why...

See on Amazon

Browse more on Amazon:The Righteous MindInfluenceThe Age of Surveillance Capitalism

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Live-tested eBay searches with available results related to this page.

Shop location

UsingUSA USA USAUKCanadaAustraliaIreland

Selected fromtechnology poster oneBay.co.uk.

Current eBay listing

RETRO TECHNOLOGY Framed Wall Art Poster Canvas Print Picture

GBP 8.99 | Shipping GBP 4.95NewSeller: luxuryprintz (99.8% positive)Ships from: GB

View listing on eBay

Current eBay listing

Aihrgdesign A Vintage Technology Po Framed Wall Art Poster Canvas Print Picture

GBP 29.97 | Shipping GBP 4.95NewSeller: artnoire (99.5% positive)Ships from: GB

View listing on eBay

Current eBay listing

aerospace technology or space engin Framed Wall Art Poster Canvas Print Picture

GBP 29.95 | Shipping GBP 4.95NewSeller: silkframe (99.7% positive)Ships from: GB

View listing on eBay

Current eBay listing

Ahead Of Technology Framed Art Prin Framed Wall Art Poster Canvas Print Picture

GBP 29.95 | Shipping GBP 4.95NewSeller: silkframe (99.7% positive)Ships from: GB

View listing on eBay

Browse more oneBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

1. Source: nature.com
Link:https://www.nature.com/articles/s41467-025-61345-5

Published: July 1, 2025

2. Source: nature.com
Title: On the conversational persuasiveness of GPT-4 | Nature Human Behaviour
Link:https://www.nature.com/articles/s41562-025-02194-6

3. Source: arxiv.org
Title: arXiv Persuasion with Large Language Models: a Survey
Link:https://arxiv.org/abs/2411.06837

Published: November 11, 2024

4. Source: arxiv.org
Title: arXiv Measuring and Improving Persuasiveness of Large Language Models
Link:https://arxiv.org/abs/2410.02653

Published: October 3, 2024

5. Source: arxiv.org
Link:https://arxiv.org/abs/2406.17753

6. Source: www-prod.media.mit.edu
Link:https://www-prod.media.mit.edu/projects/deceptive-ai-systems/overview/

7. Source: nature.com
Link:https://www.nature.com/articles/s41598-025-30783-y

8. Source: nature.com
Title: A I chatbots can persuade voters to change their minds
Link:https://www.nature.com/articles/d41586-025-03733-x

9. Source: nature.com
Title: Persuading voters using human–artificial intelligence dialogues | Nature
Link:https://www.nature.com/articles/s41586-025-09771-9.pdf

10. Source: nature.com
Title: A I is more persuasive than people in online debates
Link:https://www.nature.com/articles/d41586-025-01599-7

11. Source: nature.com
Title: How AI sources can increase openness to opposing views | Scientific Reports
Link:https://www.nature.com/articles/s41598-025-00791-z

12. Source: nature.com
Link:https://www.nature.com/collections/badbafeeca

13. Source: eprints.whiterose.ac.uk
Link:https://eprints.whiterose.ac.uk/id/eprint/241702/

Additional References

14. Source: gsb.stanford.edu
Link:https://www.gsb.stanford.edu/faculty-research/publications/llm-generated-messages-can-persuade-humans-policy-issues

15. Source: youtube.com
Title: Can AI Simulate Human Behavior? This Breakthrough Changes Everything
Link:https://www.youtube.com/watch?v=E5sBqw2if0M

16. Source: youtube.com
Title: Role of Large Language Models in Human-AI Interaction: A Critical Appraisal
Link:https://www.youtube.com/watch?v=rjZUBeDd1Eg

17. Source: par.nsf.gov
Link:https://par.nsf.gov/biblio/10438290-transparency-enhances-positive-perceptions-social-artificial-intelligence

18. Source: papers.ssrn.com
Link:https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7013259

Published: July 20, 2026

19. Source: youtube.com
Title: LLMs Prove More Persuasive Than Humans
Link:https://www.youtube.com/watch?v=GWTazDQVxrw

20. Source: openreview.net
Link:https://openreview.net/forum?id=R7FHdsZv1w

21. Source: aisi.gov.uk
Link:https://www.aisi.gov.uk/blog/how-do-ai-models-persuade-exploring-the-levers-of-ai-enabled-persuasion-through-large-scale-experiments

22. Source: philpapers.org
Link:https://philpapers.org/rec/HACCTP

23. Source: doaj.org
Link:https://doaj.org/article/6e0c9180db8d49b988580cabf618ee31