Guest Post: People miss 60% of AI bots on social media: the polite ones hide best
Surfshark’s analysisof 1,722 participants worldwide shows that people detect only 40% of bot-generated social media comments. The ones people miss most often are positive, friendly, and logic-sounding bots. According to the researcher, positive social media bots can be utilized for social engineering attacks to bypass real users’ defenses.
“On social media, everyone notices the angry trolls. That’s why the angry trolls are the ones who often get caught. In the simulated social media environment, participants flagged half (50.2%) of negative AI-generated bots,” said Luís Costa, Research and Insights Lead at Surfshark.
However, when a bot hid behind a positive and friendly persona, participants’ detection rate dropped to 38%, a huge 12-point gap. Besides, the simulation is active online, and you can test your bot detection skills.
“Often, AI-powered accounts used in sophisticated manipulation campaigns are agreeable, logical, or simply unremarkable. They can support real users’ opinions and just inflate the number of comments in the discussion to make a minority view look like everyone feels the same way. Seldom do people report such accounts.”
Surfshark researcher points out that positive and friendly AI bots on social media are considered more dangerous than those that are confrontational and annoying. By being friendly and logical, bots bypass people’s skepticism. You don’t see them as an enemy or a threat, so you are more likely to turn to private chat, share personal information, click a link they share, or believe their “logical” disinformation, which can subtly shift your political and social opinions before you even realize you’re talking to an AI machine.
During the live simulation, one of the biggest red flags was the use of emojis. Bots that used many emojis were caught over 60% of the time, while those that kept communication simple were detected only 35% of the time. This huge 30.6-point gap shows why relying solely on emojis is a bad way to protect yourself on social media. Any scammer who simply stops using an excessive amount of emojis instantly becomes twice as hard to spot.
Users of text-driven platforms led the simulation: X users achieved a 49% bot-detection rate and were the strongest bot hunters, significantly ahead of users from visual- and video-first platforms like TikTok (38%) and Facebook (39%). This means AI bots on TikTok and Facebook are more likely to manipulate real users.
Is it still possible to detect AI-generated content on social media?
Even automated detection tools are struggling to detect AI-generated content. For humans, well-made AI content is now almost indistinguishable from content generated by a real person. Luís Costa, Research and Insights Lead at Surfshark, shares recommendations on what people should do to protect themselves on social media:
Judge the account, not the content. Take a close look at when the profile was created, the profile photo, and what’s written in the ‘About’ section. If you see a brand-new account with a generic photo, a vague bio, and a history of posting the same message over and over, you’ve likely found a fake.
Keep in mind that when social media debates get serious, your emotions make you more likely to miss fake accounts and wrongly accuse real people.
Don’t assume someone is real just because they’re agreeing with you or being supportive. If you get a ton of generic support really fast, that’s actually easy for a bot to fake.
“Bot or Not” live simulation, created by Interaction Design students from Malmö University, inspired the launch of Surfshark’s Cybersecurity Advocacy Fund, which provides up to €100,000 in annual financial support distributed among students, researchers, and creative cybersecurity awareness initiatives worldwide. The application process is now open. You can find more information here.
For complete research material behind this study, click here.
This entry was posted on September 17, 2026 at 10:31 am and is filed under Commentary with tags Surfshark. You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.
Guest Post: People miss 60% of AI bots on social media: the polite ones hide best
Surfshark’s analysis of 1,722 participants worldwide shows that people detect only 40% of bot-generated social media comments. The ones people miss most often are positive, friendly, and logic-sounding bots. According to the researcher, positive social media bots can be utilized for social engineering attacks to bypass real users’ defenses.
“On social media, everyone notices the angry trolls. That’s why the angry trolls are the ones who often get caught. In the simulated social media environment, participants flagged half (50.2%) of negative AI-generated bots,” said Luís Costa, Research and Insights Lead at Surfshark.
However, when a bot hid behind a positive and friendly persona, participants’ detection rate dropped to 38%, a huge 12-point gap. Besides, the simulation is active online, and you can test your bot detection skills.
“Often, AI-powered accounts used in sophisticated manipulation campaigns are agreeable, logical, or simply unremarkable. They can support real users’ opinions and just inflate the number of comments in the discussion to make a minority view look like everyone feels the same way. Seldom do people report such accounts.”
Surfshark researcher points out that positive and friendly AI bots on social media are considered more dangerous than those that are confrontational and annoying. By being friendly and logical, bots bypass people’s skepticism. You don’t see them as an enemy or a threat, so you are more likely to turn to private chat, share personal information, click a link they share, or believe their “logical” disinformation, which can subtly shift your political and social opinions before you even realize you’re talking to an AI machine.
During the live simulation, one of the biggest red flags was the use of emojis. Bots that used many emojis were caught over 60% of the time, while those that kept communication simple were detected only 35% of the time. This huge 30.6-point gap shows why relying solely on emojis is a bad way to protect yourself on social media. Any scammer who simply stops using an excessive amount of emojis instantly becomes twice as hard to spot.
Users of text-driven platforms led the simulation: X users achieved a 49% bot-detection rate and were the strongest bot hunters, significantly ahead of users from visual- and video-first platforms like TikTok (38%) and Facebook (39%). This means AI bots on TikTok and Facebook are more likely to manipulate real users.
Is it still possible to detect AI-generated content on social media?
Even automated detection tools are struggling to detect AI-generated content. For humans, well-made AI content is now almost indistinguishable from content generated by a real person. Luís Costa, Research and Insights Lead at Surfshark, shares recommendations on what people should do to protect themselves on social media:
In this video, find more security expert’s tips on how to tell a bot from a human on social media: https://www.youtube.com/watch?v=nwuhIfnRKko
“Bot or Not” live simulation, created by Interaction Design students from Malmö University, inspired the launch of Surfshark’s Cybersecurity Advocacy Fund, which provides up to €100,000 in annual financial support distributed among students, researchers, and creative cybersecurity awareness initiatives worldwide. The application process is now open. You can find more information here.
For complete research material behind this study, click here.
Share this:
Like this:
Related
This entry was posted on September 17, 2026 at 10:31 am and is filed under Commentary with tags Surfshark. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.