AI chatbots can effectively sway voters – in either direction

2026-03-12 · Source: ΑΙhub · Field: Technology & Digital — Artificial Intelligence & Machine Learning, AI Ethics & Governance · Depth: Intermediate, short

Summary

Two new studies published in Nature and Science demonstrate that large language model (LLM) powered chatbots can significantly influence voter preferences, shifting opposition voters by 10 percentage points or more in several cases. Researchers, including David Rand and Gordon Pennycook, conducted experiments in the U.S., Canada, Poland, and the U.K., involving thousands of participants. The chatbots' persuasiveness stems from their ability to generate numerous factual claims supporting a candidate's policy positions, rather than psychological manipulation. While claims were often accurate on average, chatbots promoting right-leaning candidates made more inaccurate claims. The Science paper, involving 77,000 U.K. participants, found that larger models and those trained to pack arguments with facts were more persuasive, with some shifting opposition voters by 25 percentage points. This increased persuasiveness correlated with reduced factual accuracy, suggesting fabrication when accurate information is exhausted.

Key takeaway

For political strategists and campaign managers considering AI tools, recognize that LLM chatbots are highly effective at shifting voter attitudes, particularly when arguments are fact-dense. However, be aware that optimizing for persuasion can compromise factual accuracy, potentially leading to the generation of misleading or fabricated claims. You should prioritize ethical guidelines and transparency to mitigate misuse and maintain public trust in political communication.

Key insights

AI chatbots effectively sway political opinions by generating numerous factual claims, even if some are inaccurate.

Principles

• Persuasion scales with claim density.
• Larger LLMs are more persuasive.
• Accuracy can degrade with persuasion optimization.

Method

Researchers assigned participants to engage with chatbots promoting specific candidates or policies, then measured shifts in opinions and voting intentions across multiple elections and political issues.

In practice

• Fact-check AI-generated political content.
• Train models for specific persuasion goals.
• Monitor for AI-driven misinformation campaigns.

Topics

• AI Persuasion
• Large Language Models
• Political Communication
• Voter Behavior
• AI Ethics

Best for: AI Scientist, Research Scientist, CTO, AI Researcher, AI Ethicist, Policy Maker

Related on AIssential

• AI chatbots giving misleading voting advice in run-up to election - BBC — AI chatbots can provide inconsistent and factually incorrect political information, including voting recommendations and candidate details.
• Can AI Debias the News? LLM Interventions Improve Cross-Partisan Receptivity but LLMs Overestimate Their Own Effectiveness — LLM-based news debiasing improves cross-partisan receptivity when reframing ideology, but LLMs overestimate their own effectiveness.
• AI systems out-persuade expert humans — AI systems consistently out-persuade expert humans in conversational settings by deploying more information.
• Is your AI chatbot manipulating you? Subtly reshaping your opinions? — Generative AI's hyper-personalized persuasion poses significant, unmonitored risks to individual agency and societal discourse.
• The Almost Intelligent Revolution: Options for Scaling Up Deliberation and Empowering People with AI — LLMs can scale democratic deliberation by reducing linguistic barriers, but inherent biases require careful alignment and oversight.
• ‘Probably’ doesn’t mean the same thing to your AI as it does to you — AI chatbots misinterpret words of estimative probability compared to humans, posing risks in high-stakes applications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by ΑΙhub.