What does this research mean for the field?

LLM-driven debates effectively reduce belief in health-related conspiracy theories, but this reduction is significantly stronger when the chatbot is perceived as a human rather than an AI due to differences in perceived neutrality. Novelty: ClaimNovelty.INCREMENTAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To investigate whether Large Language Model (LLM) debates reduce health-related conspiracy beliefs and evaluate how AI versus human source attribution influences this reduction.

August 19, 2026Scientific Reports0 citationsOpen Access

Chatbots Reduce Health Conspiracy Beliefs Despite Artificial Intelligence Labeling

Chatbots reduce health-related conspiracy beliefs not because of but despite being perceived as AI

View Full Paper

Ask AI

Bookmark

Authors

PBPaul Ballot YSYana van de Sande HSHanna Schraffenberger

Discussion

Member takes

Overview

Randomized trial demonstrates that conversational AI reduces COVID-19 conspiracy beliefs in adults, indicating perceived source neutrality enhances debiasing effectiveness.

Key Points

To investigate whether Large Language Model (LLM) debates reduce health-related conspiracy beliefs and evaluate how AI versus human source attribution influences this reduction.
Randomized online experiment with 554 participants assigned to a control condition or a debate over a COVID-19 conspiracy theory.
Debates were conducted using an LLM that was either labeled as an artificial intelligence or as a human partner.
Participants debating an AI-labelled LLM showed a 7.88 percentage point decrease in conspiracy confidence compared to control (d = 0.63, p < .001).
Participants debating a human-labelled LLM experienced a 13.76 percentage point larger confidence drop compared to control (d = 0.87, p < .001).
The enhanced debiasing effect in the human-labelled group was mediated by the perceived neutrality of the source.

Cite This Study

Ballot et al. (2026) studied this question.

synapsesocial.com/papers/6a8562eb03308d306e2d5ec9 https://doi.org/10.1038/s41598-026-66242-5

View Full Paper

Ask AI

Bookmark

August 19, 2026Scientific Reports0 citationsOpen Access

Chatbots Reduce Health Conspiracy Beliefs Despite Artificial Intelligence Labeling

Chatbots reduce health-related conspiracy beliefs not because of but despite being perceived as AI

View Full Paper

Ask AI

Bookmark

Authors

PBPaul Ballot YSYana van de Sande HSHanna Schraffenberger

Discussion

Member takes

Overview

Randomized trial demonstrates that conversational AI reduces COVID-19 conspiracy beliefs in adults, indicating perceived source neutrality enhances debiasing effectiveness.

Key Points

To investigate whether Large Language Model (LLM) debates reduce health-related conspiracy beliefs and evaluate how AI versus human source attribution influences this reduction.
Randomized online experiment with 554 participants assigned to a control condition or a debate over a COVID-19 conspiracy theory.
Debates were conducted using an LLM that was either labeled as an artificial intelligence or as a human partner.
Participants debating an AI-labelled LLM showed a 7.88 percentage point decrease in conspiracy confidence compared to control (d = 0.63, p < .001).
Participants debating a human-labelled LLM experienced a 13.76 percentage point larger confidence drop compared to control (d = 0.87, p < .001).
The enhanced debiasing effect in the human-labelled group was mediated by the perceived neutrality of the source.

Cite This Study

Ballot et al. (2026) studied this question.

synapsesocial.com/papers/6a8562eb03308d306e2d5ec9 https://doi.org/10.1038/s41598-026-66242-5

View Full Paper

Ask AI

Bookmark