AI can persuade people to take political actions.
4 affiliations
- 1Oxford Internet Institute, University of Oxford, Oxford OX2 6GG, United Kingdom.
- 2UK AI Security Institute, London SW1A 2AS, United Kingdom.
- 3Department of Sociology, Stanford University, Stanford, CA 94305.
- 4The Department of Psychological and Behavioural Science, London School of Economics and Political Science, London WC2A 2AE, United Kingdom.
Abstract
There is substantial concern about the ability of advanced AI to influence people's behavior. A rapidly growing body of research has found that AI can produce large persuasive effects on people's attitudes, but whether AI can persuade people to take consequential real-world actions has remained unclear. In two large preregistered experiments ([Formula: see text] responses from 14,779 people), we used conversational AI models to persuade participants on a range of attitudinal and behavioral outcomes, including signing real petitions and donating money to charity. We found sizable AI persuasion effects on these behavioral outcomes (e.g. [Formula: see text]19.7 percentage points on petition signing). However, we observed no evidence of a correlation between AI persuasion effects on attitudes and effects on behavior. Moreover, we replicated prior findings that information provision drove effects on attitudes, but found no such evidence for our behavioral outcomes. In a test of eight behavioral persuasion strategies, the strategy focused on information provision produced the smallest behavioral effect, though differences among the eight were small. Taken together, these results suggest that previous findings relying on attitudinal outcomes may generalize poorly to behavior, and therefore risk substantially mischaracterizing the real-world behavioral impact of AI persuasion.
Related
- Kicking politics: Towards a framework for detecting political influence in online communities
- Quantification of Enzyme–Inhibitor Pharmacological Actions
- Glucocorticoid signalling exerts distinct lineage-permissive and dose-dependent disruptive actions during human osteoblast differentiation
- Decoding protein signaling networks in microvascular angina: the multi-target biochemical actions of traditional Chinese medicine
- Effectiveness and safety of coenzyme Q10 in people with heart failure with preserved ejection fraction (HFpEF): a systematic review and meta-analysis
BibTeX
@article{hackenburg2026persuade,
title = {{AI can persuade people to take political actions.}},
author = {Kobi Hackenburg and Luke Hewitt and Caroline Wagner and Christopher Summerfield and Ben M Tappin},
journal = {Proceedings of the National Academy of Sciences of the United States of America},
year = {2026},
month = oct,
doi = {10.1073/pnas.2611368123},
url = {https://europepmc.org/article/MED/42886696},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - AI can persuade people to take political actions.
AU - Kobi Hackenburg
AU - Luke Hewitt
AU - Caroline Wagner
AU - Christopher Summerfield
AU - Ben M Tappin
PY - 2026
DA - 2026/10/16
JO - Proceedings of the National Academy of Sciences of the United States of America
DO - 10.1073/pnas.2611368123
UR - https://europepmc.org/article/MED/42886696
AB - There is substantial concern about the ability of advanced AI to influence people's behavior. A rapidly growing body of research has found that AI can produce large persuasive effects on people's attitudes, but whether AI can persuade people to take consequential real-world actions has remained unclear. In two large preregistered experiments ([Formula: see text] responses from 14,779 people), we used conversational AI models to persuade participants on a range of attitudinal and behavioral outcomes, including signing real petitions and donating money to charity. We found sizable AI persuasion effects on these behavioral outcomes (e.g. [Formula: see text]19.7 percentage points on petition signing). However, we observed no evidence of a correlation between AI persuasion effects on attitudes and effects on behavior. Moreover, we replicated prior findings that information provision drove effects on attitudes, but found no such evidence for our behavioral outcomes. In a test of eight behavioral persuasion strategies, the strategy focused on information provision produced the smallest behavioral effect, though differences among the eight were small. Taken together, these results suggest that previous findings relying on attitudinal outcomes may generalize poorly to behavior, and therefore risk substantially mischaracterizing the real-world behavioral impact of AI persuasion.
SN - 0027-8424
SN - 1091-6490
ER -