Experimental AI chatbots significantly reduce belief in anti-Semitic conspiracy theories, new ADL-supported study shows
NEW YORK CITY –A study recently released by independent researchers supported by ADL (the Anti-Defamation League) found that short dialogues with a large language model (LLM) programmed to engage with believers of certain anti-Semitic conspiracy theories reduced their belief in anti-Semitic conspiracy theories by 16 percent and increased favorability toward Jews by 25 percent among initially unfavorable participants.
Successful Intervention
For this research, experts carried out an experiment involving 1,224 U.S. adults who endorsed at least one of six anti-Semitic conspiracy theories. Participants engaged in a brief conversation about their specific conspiracy belief with an AI chatbot (Claude 3.5 Sonnet) programmed to debunk antisemitic conspiracy theories.
“For millennia, anti-Semitic conspiracy theories have fueled hatred and violence against Jewish people,” said Jonathan A. Greenblatt, ADL CEO and national director. “At ADL, we are constantly innovating to find new ways to counter this age-old hatred with modern technology —it’s now very encouraging to see that even deeply entrenched prejudice can be effectively challenged by promising new tools that could reach millions at scale.”
The study also found that the effects persisted over time, with roughly 50 percent of the initial decrease caused by the intervention still evident over a month later. The intervention worked across different types of anti-Semitic conspiracy theories, from claims about Jewish control of media and government to Holocaust denial and theories about Jewish responsibility for COVID-19.
“What’s remarkable about these findings is that factual debunking works even for conspiracy theories with deep historical roots and strong connections to identity and prejudice,” said David G. Rand, professor of information science, marketing, and psychology at Cornell University and the study’s senior author. “Our artificial intelligence debunker bot typically doesn’t rely on emotional appeals, empathy-building exercises, or anti-bias tactics to correct false beliefs. It mostly provides accurate information and evidence-based counterarguments, demonstrating that facts still matter in changing minds.”
Challenge To Assumption
A 2024 ADL study revealed widespread and increasing endorsement of anti-Jewish trope beliefs among the general population. Nearly a quarter of the initial survey sample for this new report endorsed three or more anti-Semitic conspiracy theories, with the most commonly believed theory involving powerful Jewish families manipulating world events. Moreover, this new study —as with previous ADL research —also found a unique correlation between anti-Semitic conspiracy beliefs and prejudice against Jews—stronger than the relationship between general conspiracy thinking and anti-Semitism.
The findings challenge the assumption that anti-Semitic beliefs tied to identity and centuries of historical prejudice are immune to factual correction. Previous research by the same team showed similar results for classic conspiracy theories, but this study is the first to demonstrate that AI-powered debunking can reduce both anti-Semitic conspiracy beliefs and the prejudice they fuel.
“This research shows the potential is enormous, the next step is figuring out how to leverage this to meet people where they are and get them to use these tools,” said Matt Williams, vice president of the ADL Center for Antisemitism Research. “We must explore strategies like integration into search engines and social media platforms, recommendations from trusted messengers and public awareness campaigns.”
Data for this study was collected in January 2025 with follow-up data gathered in February 2025.
The full research report is available here.
