The rise of large language models (LLMs) such as ChatGPT, Llama, Claude, and Grok has had the unfortunate effect of turbocharging existing efforts at flooding social media and the internet in general with clickbait and spam. Sometimes, however, this results in mistakes that make spam networks more obvious, such accidentally posting an LLM prompt rather than the LLM’s output to one or more social media accounts. In one such recent incident, several spam accounts on both Bluesky and X posted prompts beginning with “Give me a title similar to the curious content that makes users click to see…”, accompanied by links to a pair of plagiarized “news” sites.
The two websites in question, luxuryhousezone(dot)com and 3sblog(dot)com, feature a wide variety of news articles, most (possibly all) of which were plagiarized from various popular news sites and blogs. In many cases, the plagiarized sites simply lifted text verbatim from the original articles, without even bothering to reword it using a large language model. Images and embedded videos were generally duplicated as well. The sites feature content in a mix of languages, although English is predominant.
Each website has a dedicated network of Bluesky accounts promoting it. The network promoting luxuryhousezone(dot)com is made up of accounts masquerading as individual people. These accounts use stolen profile images, and some have repetitive biographies, which were flagged by the process described in this previous post. This network presently consists of 574 accounts, although at least 822 have been created; 248 have been suspended by Bluesky.
The other network of Bluesky accounts, which shares links to 3sblog(dot)com, is made up of a mix of accounts presenting as people and accounts decorated to look like news organizations. The bogus news organization accounts frequently reuse the same banner images across multiple accounts. 135 accounts belonging to this network have been detected thus far, 57 of which are still active and 78 of which are presently suspended.
Combined, these two spam networks consist of a total of 957 accounts, all created on December 1st, 2024, or later. 631 of these accounts are presently online; the remaining 326 have been suspended by Bluesky. The accounts were created in large batches, with upwards of 70 accounts created in an hour at multiple points in time. The account creations tend to switch batch and forth between the two networks.
Thus far, the Bluesky accounts in each network exclusively post links to (plagiarized) articles on luxuryhousezone(dot)com and 3sblog(dot)com. Each account links one site or the other; thus far, none has linked both. The articles cover a wide range of topics, although U.S. politics is a recurring theme. In some of the newer posts, the links are redirected through buff.ly, suggesting that the accounts in the spam networks have been automated using Buffer.
The content is highly repetitive, with each article linked by multiple spam accounts and many articles linked by dozens. The most frequently linked article to date, with 41 posts on Bluesky, bears the title “BREAKING NEWS: Candace Owens SPEAKS OUT, demanding that the NFL cut ties with Taylor Swift, claiming that the pop star’s erratic behavior and statements go beyond acceptable limits… see more:”. In most cases, the text of the post matches the title of the article, with the exception of the aforementioned glitched posts where an LLM prompt was posted instead.
This activity is not confined to Bluesky; networks promoting both websites were created on X (formerly known as Twitter) back in October 2024. The networks on X are smaller, with 68 accounts linking to luxuryhousezone(dot)com and 14 accounts linking to 3sblog(dot)com. Unlike the Bluesky accounts, the X accounts do not limit themselves to promoting the pair of plagiarized news websites, as most post photos and videos as well (frequently NSFW). In another difference from the Bluesky spam accounts, one account (@JohnBaker426149) links both sites.
As mentioned earlier, and as is often the case with spam accounts, both the Bluesky and X accounts in these networks use stolen profile photos. TinEye reverse image search revealed multiple prior occurrences of most of the photos of actual human beings used as avatars by the spam accounts. (Some of the accounts use either default profile images or logos of nonexistent news organizations as avatars, rather than photos of humans.)
The people at Bluesky should hire you! Great work finding this garbage. Not sure who benefits from it.