The networks of fake accounts that post pro-Russia and pro-Putin comments on social media and demonize Ukraine have become more sophisticated thanks to artificial intelligence technologies. They have begun to write more unique messages and follow human routines, but they sometimes fail.
The networks of fake accounts that post pro-Russia and pro-Putin comments on social media and demonize Ukraine have become more sophisticated thanks to artificial intelligence technologies. They have begun to write more unique messages and follow human routines, but they sometimes fail.
This is stated in a study by the publication «Ukrainska Pravda». It takes into account the analysis of 3,600 bots in Telegram channels of the occupied territories of Ukraine, which over a year and a half wrote at least 316,000 comments in support of the Russian Federation and the occupation.
The researchers counted 69 different topics that the bots wrote about. One bot could work in dozens of channels throughout the day and leave comments on radically different topics.
«For example, a user under the nickname nieucqanoberr wrote 1,449 comments in 65 channels on 40 different topics during May 11, 2024. That is, on average, he left a comment every 36 seconds. Telegram eventually blocked him due to excessive activity,» the study says.
It has become harder for social media algorithms and moderators to detect bots, as Russian bots are more active during the day and less active at night to better mimic the behavior of real users. Bot farms have also started writing more unique posts thanks to AI.
If in Telegram channels of the occupied territories of Ukraine only a third of comments were unique, then during the elections in Moldova in 2025, AI ensured 95% uniqueness. Bots used templates less and adapted propaganda to the context of the news. Sometimes this led to curiosities.
For example, under a post about rain in one of the Moldovan cities, bots left more than a dozen comments saying that Maia Sandu had ruined the weather in the country.
To write non-standard comments, the bot farm algorithm monitors and analyzes the information space of the social network. When a new publication appears, the algorithm determines which topic from the list is most relevant to it, and the bots begin to publish adapted comments with the necessary narratives.
This process requires minimal human involvement, but the most difficult task for organizers is to teach the algorithm to understand which posts to promote which messages under. However, the algorithm learns quickly from huge amounts of data. So now bots make fewer and fewer mistakes.
