What is it about?
Generative AI tools like ChatGPT and image generators make it easy to produce large amounts of text and images. We wanted to know whether people who spread hate online are actually using these tools, and whether doing so makes hate worse. We studied 4chan's /pol/ board, a largely unmoderated forum known for extreme content, and compared two long-standing forms of hate: hate against Indian people and antisemitism. We first tested several AI-content detectors to find the most reliable ones, then used them to check more than 31,000 long hateful posts and about 50,000 hateful images.
Featured Image
Photo by Joshua Hoehne on Unsplash
Why is it important?
AI-generated hateful posts have grown steadily since late 2022, reaching 1.42% of long posts by 2025. But the picture differs between the two targets. For anti-Indian hate, AI use rises together with the use of slurs, nearly doubles during spikes of hate, and AI-written posts are about 56% more toxic than human-written ones. For antisemitic hate, we saw no such link. This suggests AI is used mostly in organized, campaign-style hate. Hate campaigns also tend to begin about three months before AI use picks up, so AI seems to speed up existing hate movements rather than create new ones. For platforms and policymakers, watching for emerging hate campaigns may be more effective than trying to spot AI-generated content alone.
Perspectives
Much public debate assumes AI will flood the internet with new kinds of hate. Our results point to a more specific risk: AI acts as an amplifier for people who are already running hate campaigns. I hope this helps focus moderation efforts where they matter most.
Bo Shao
CISPA Helmholtz Center for Information Security
Read the Original
This page is a summary of: Does AI Amplify Online Hate- A Multi-Modal Measurement Study on 4chan's /pol/, November 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3830454.3846752.
You can read the full text:
Contributors
The following have contributed to this page
