Surfshark Study Finds Polite AI Bots Harder to Spot on Social Media
Summary
In a Surfshark study of 1,722 people, participants identified 40% of AI-generated social media comments. Positive, agreeable bots were harder to spot than negative ones, raising concerns about manipulation and privacy.
Key points
- Participants identified 40% of the AI-generated comments overall; they detected 50.2% of negative bots and 38% of positive, agreeable bots.
- Serious topics led to more missed bots and more real users wrongly accused of being bots than lighter topics.
- Bots using many emojis were detected more than 60% of the time, while plain-language bots were detected 35% of the time.
- Threads users had the highest detection rate, but Surfshark noted the sample was small; X users followed, while TikTok and Facebook users performed less well.
- Detection declined with age and was lowest among participants over 50, who also had the weakest accuracy in identifying real people.
- The article says researchers consider agreeable bots a privacy and security concern because people may trust them, share personal information, or click links without realizing they are bots. These are potential risks, not reported outcomes of the test.
Article Details
- Publisher
- Help Net Security
- Scope
- Surfshark study of participants’ ability to distinguish human comments from AI-generated comments in a social media setting
- Sample Size
- 1,722 participants worldwide
- Key Statistics
- Participants correctly identified 40% of AI-generated bots overall.
- Participants flagged 50.2% of negative AI-generated bots, compared with 38% of positive, agreeable bots.
- Bots using emojis heavily were caught over 60% of the time; bots using plain language were detected 35% of the time.
- Threads users had the highest detection rate, but Surfshark noted the sample was small.
People
Products
FacebookThreads and X users performed better than TikTok and Facebook users, while older participants struggled most.ThreadsThreads and X users performed better than TikTok and Facebook users, while older participants struggled most.TikTokThreads and X users performed better than TikTok and Facebook users, while older participants struggled most.XThreads and X users performed better than TikTok and Facebook users, while older participants struggled most.