You spammed it magnificently!
Come for the repetitive compliments, stay for the AI-generated avatars
If you recently had a post go viral on X (formerly Twitter), you may have had a bunch of total strangers show up to reply to your post with extremely similar compliments. These replies, which simply consist of statements such as “You did it magnificently!”, “You did it perfectly!”, and “You did it superbly!” repeated ad nauseam, are from a spam network that appears to have kicked into high gear in early June 2024. This network primarily spews its spammy output in response to viral posts about AI or cryptocurrency, although it regularly branches out to other topics as well. Many of the spam accounts use artificially generated faces as profile images, while others use stolen photos that have also been used by previous spam networks.
This spam network consists of at least 442 X accounts, created between June 2023 and June 2024, mostly in the latter half of that period. These accounts were found by searching for replies containing the text “You did it magnificently!” and manually removing false positives. Many of the accounts in the network have a peculiar naming scheme, wherein the display name and the handle are similar, but the handle has fewer consonants and more vowels. Examples include the account with the display name “Purpureal”, which has the handle @Puriuouaaal, and “Roy Lam”, which has the handle @RoooLaaaoooo.
258 of the 442 accounts in the network (58.4%) use StyleGAN-generated faces as profile images, with the remainder using a mix of plagiarized photos and default avatars. Most of these artificially generated faces have neutral backgrounds, and many have a “StyleGAN 2 (Karras et al.)” watermark in the lower right-hand corner of the image. As with all unmodified StyleGAN-generated faces, the placement of the major facial features, particularly the eyes, is extremely consistent from image to image. Both the use of artificially generated faces and the vowel-heavy naming scheme are characteristics that this network has in common with two previous spam networks, which indicates that all three networks were created using similar methods, and may be operated by the same people.
One benefit of using AI image generation tools such as StyleGAN to generate fake profile photos is that one can easily generate an endless number of unique images. Despite this, the operators of this network have in several instances used the same artificially generated face as the avatar for multiple accounts. The same is true of the plagiarized photos used by the network, several of which turn up as profile images for more than one account. Some of the plagiarized photos were also used as avatars by one of the network’s potential predecessors.
A significant portion of this network’s content has been archived in JSON format on Wayback Machine, making it possible to study this network’s posts despite the removal of the free X/Twitter API. A random sample of 15000 archived posts (out of roughly 60000 total) was downloaded and used for this analysis. Most of the network’s content is replies, and most of these replies are repetitive compliments such as “You did it superbly!”, “You did it perfectly!”, “You were exceptional!”, and the aforementioned “You did it magnificently!”, each of which turned up hundreds of times in the dataset analyzed. Note that since the dataset studied comprises at most a quarter of the network’s total output, the real numbers are likely at least four times those shown in the table below.
In addition to the repetitive replies, the network also reposts an eclectic variety of content. The posts replied to and reposted by the network are a mix of viral posts from legitimate accounts (with @elonmusk and @cenyk1230 being particularly popular) and posts from other spam accounts. AI and cryptocurrency are recurring themes of the posts amplified by the network, but everything from news to sports to politics to porn turns up as well.
You substacked it magnificently! :D
You were exceptional!!!!