Abstract
Abstract
Social media platforms mediate how billions form opinions and engage with public discourse. As autonomous AI agents increasingly participate in these spaces, understanding their behavioral fidelity becomes critical for platform governance and democratic resilience. Previous work demonstrates that LLM-powered agents can replicate aggregate survey responses, yet few studies test whether agents can predict specific individuals’ reactions to specific content. This study benchmarks LLM-based agents’ accuracy in predicting human social media reactions (like, dislike, comment, share, no reaction) across 120,000 + unique agent-persona combinations derived from 1,511 Serbian participants and 27 large language models. In Study 1, agents achieved 70.7% overall accuracy, with LLM choice producing a 13%-point performance spread. Study 2 employed binary forced-choice (like/dislike) evaluation with chance-corrected metrics. Agents achieved Matthews Correlation Coefficient (MCC) of 0.29, indicating genuine predictive signal beyond chance. However, conventional text-based supervised classifiers using TF-IDF representations outperformed LLM agents (MCC of 0.36), indicating that the predictive signal derives from semantic text content rather than from any capacity for individualized behavioral simulation. The genuine but modest predictive validity of zero-shot persona-prompted agents suggests that, while current accuracy is insufficient for precise individual targeting, the capacity to predict reactions at rates above chance warrants attention in discussions of AI-driven influence and social simulation methodology. The advantage of zero‑shot agents is that they require no task‑specific training, which makes large‑scale deployment easy across diverse contexts, including election campaigns and mass‑scale manipulation. Limitations include single-country sampling. Future research should explore multilingual testing and fine-tuning approaches.