A new experiment from cybersecurity company Surfshark suggests that even people who consider themselves savvy online users are struggling to tell AI bots apart from real humans on social media.
Of the 710 participants who took part in the study carried out with master’s students from Malmö University, only 53% correctly identified more bots than they misidentified humans. This means that nearly half (47%) failed the task altogether.
Recent industry estimates suggest bot-driven amplification now accounts for around 23% of political discourse on X during election seasons.
Surfshark’s own earlier research found that major platforms remove more than 6.3 billion fake accounts each year, roughly 47 times the number of babies born worldwide annually.
Even the best VPN cannot make you better at recognising an AI-written comment, and that is exactly the gap this experiment is trying to highlight.
The “Bot or Not” simulation puts you in the seat of a content moderator and asks one simple question: Can you really still trust your own instincts when you scroll?
Inside Surfshark’s “Bot or Not” experiment
The “Bot or Not” game is a timed, interactive simulation built by Interaction Design master’s students at Malmö University for the UNFOLD exhibition during Milan Design Week.
Players are dropped into a simulated social media comment section and given 120 seconds to spot 10 bot-written comments across four discussion topics.
Two of those topics were deliberately “cold,” meaning low in emotional charge: data centres and the perennial pineapple-on-pizza debate. The other two were “hot” and politically loaded: immigration and women’s rights. The contrast between the four was where the most revealing data appeared.
When participants discussed data centres, they identified 71% of the bots with a 76% accuracy rate, the strongest result in the study. Pineapple on pizza was almost as good, at 64% detection and 69% accuracy.
The moment the simulation moved into emotional territory, however, performance collapsed.
On immigration, detection fell to 54% and accuracy to 63%. On women’s rights, detection crashed to just 49%, with accuracy slipping to 61%, meaning users were both missing more bots and wrongly accusing more real humans of being machines.
Who struggles most, and how to take the test
The study also points to a clear “generational cliff” at around the age of 40. Players up to age 20 were the strongest bot-hunters in the dataset, finding nearly 65% of bots with an accuracy of more than 71%. Performance held steady through the 20s and 30s, then dropped sharply for the 41 to 50 bracket, where detection fell to 42% and accuracy to 59%. Users over 50 fared only marginally better.
According to Surfshark’s Research Lead Luís Costa, the takeaway is not really about reading skills or media literacy in the traditional sense. The biggest blind spot the experiment exposed was emotion: when a debate gets heated, it effectively hijacks the mental “radar” people rely on to flag suspicious content.
To push back against automated deception, he argues, what users actually need is a cooler head and a better awareness of their own vulnerabilities, not sharper textual analysis.
The “Bot or Not” game is now publicly available at botornot.one, and anyone can play it in their browser to see how they score against the original 710 participants.
The wider point of the study is harder to shake off than the score on any individual playthrough. Bots are being produced by the billions, the technology that powers them is getting better at blending in, and our own emotional reactions are the lever they are increasingly built to pull.
A few minutes with “Bot or Not” is a quick way to find out just how often that lever is already working on you.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!