A recent study by Surfshark, a cybersecurity firm, tested how well people can spot AI-generated comments on social media. The researchers asked 1,722 participants from around the world to judge whether comments came from a human or a bot. The results show that most people struggle with this task, and the bots that fool them most often are the ones that behave politely.
The study suggests that common bot-detection cues—like aggressive or odd phrasing—are less reliable than assumed. Instead, AI systems that mimic courteous, conversational tone blend in more easily with genuine user interactions. This raises concerns for social media platforms trying to moderate automated content, as politeness may be a deliberate tactic to evade detection.
While the study does not reveal specific detection strategies, it highlights the growing sophistication of AI-generated content. For users, the takeaway is that a friendly comment is not necessarily a human one. The findings also point to a need for better tools to identify bots that adopt socially acceptable language patterns.