Adversarial Creation and Detection of AI-Generated Social Bot Content

2026-06-05

AI Summary

Current methods for detecting AI-generated content exhibit limited performance in real-world scenarios, primarily due to the scarcity of data that reflects authentic adversarial conditions. This work proposes an adversarial data construction paradigm that simulates malicious actors impersonating genuine social media users, yielding the first multilingual, cross-platform dataset of paired human- and AI-generated texts. A detector trained on this dataset substantially outperforms existing content-based social bot detection approaches on out-of-distribution real-world data, achieving, for the first time, high-accuracy identification of AI-generated content under realistic attack conditions.

Abstract

The convergence of large language models and social bots allows malicious actors to manipulate the information ecosystem by generating human-like content at scale. Existing models for detecting AI-generated content often fail in the wild, primarily due to the lack of ground-truth data. We address this gap through an adversarial methodology that models the impersonation of real social media users by malicious actors. Using this methodology, we curate a multilingual, cross-platform dataset of paired human and AI-generated messages. Training on such adversarial data yields accurate detection of AI-generated text. Our approach significantly outperforms existing models for content-based bot detection in real-world, out-of-distribution data.

Problem

• AI-generated content
• social bots
• adversarial detection
• ground-truth data
• content moderation

Innovation

• adversarial generation
• AI-generated content detection
• social bot impersonation
• multilingual dataset
• out-of-distribution detection