NASK at the 64th Annual Meeting of the Association for Computational Linguistics

Arkadiusz Modzelewski – Senior NLP Specialist, and Dr. Anna Kołos – Head of Linguistic Engineering, represented NASK – National Research Institute at the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026, CORE A*, 200 points in Poland) in San Diego, USA — one of the world’s most prestigious conferences in Natural Language Processing and Artificial Intelligence.

During the conference, they presented scientific article titled “Can AI-Generated Persuasion Be Detected? Persuaficial Benchmark and AI vs. Human Linguistic Differences”.

The work addresses an increasingly important challenge: can we distinguish AI-generated persuasive content from human-written persuasion?

In the paper, authors introduce Persuaficial, a new high-quality multilingual benchmark covering six languages (English, German, Polish, Italian, French, and Russian). Their experiments show that while overt AI-generated persuasion can often be detected, subtle persuasive content generated by LLMs significantly reduces the performance of current detection systems. They also present the first comprehensive linguistic comparison between AI-generated and human-authored persuasive texts, providing insights that can support the development of more robust and interpretable detection methods.