According to the report by NewsGuard, artificial intelligence (AI) tools are now being utilized to create “content farms,” low-quality spammy websites that generate numerous clickbait articles to maximize advertising revenue.
NewsGuard found 49 websites in April 2023 across seven different languages (Chinese, Czech, English, French, Portuguese, Tagalog, and Thai). These sites are suspected to be completely or largely generated by language models that mimic human communication. They present themselves as conventional news websites.
The sites cover topics such as politics, health, entertainment, finance, and technology, producing a high volume of content. Some publish hundreds of articles each day, but many of the articles advance false narratives. The use of AI is noticeable due to the repetitive language and bland phrases that appear throughout the content.
The high volume of advertisements on these websites indicates that they were likely created to generate revenue through programmatic ads, which are placed algorithmically across the internet and are a primary source of funding for much of the world’s media. This mirrors the approach of the first generation of content farms, which humans operated.
The source of increasingly powerful AI tools has raised concerns that they could be used to create entire news organizations, an idea that was once the subject of hypothesis by media scholars but has now become a reality.
In April 2023, They contacted 29 of 49 sites included in their analysis that listed contact information. Two of these sites confirmed the use of AI, while two did not respond to queries. Eight provided invalid email addresses, and 17 did not respond at all.
Maria Spanadoris, the owner of Famadillo.com, has published many AI-generated product reviews credited to “admin.” However, Spanadoris denied the extensive use of AI on the site, claiming that they only tested its capability on old, low-traffic articles.
Adesh Ingale, who introduced himself as the founder of GetIntoKnowledge.com, has posted clickbait articles created by AI regarding various topics, including history and science.
In response, Ingale stated that they use automation only when necessary and assured that their content is 100% fact-checked to prevent disseminating false information. He also emphasized the integration of automation software into their workflow as the world moves towards a digital and automated era. According to Ingale, all the results generated from this approach are based on authentic regional facts.
These sites typically have generic names that could easily be mistaken for legitimate publishers. Examples include Biz Breaking News, News Live 79, Daily Business Post, and Market News Reports.
The AI-generated articles on these sites are often summaries or rewrites of content from other sources. BestBudgetUSA.com, for instance, is an anonymously registered site that appears to rewrite CNN articles mainly. The website does not provide information about its ownership.
Many of the articles produced by AI on these sites contain language that gives away their machine-generated nature. For example, numerous articles on BestBudgetUSA.com include phrases that are typical of generative AI, like “I am not capable of producing 1500 words… Yet, I can provide you with a summary of the article,” followed by a link to the original CNN report.
Such phrases suggest that these sites are likely operating without significant human oversight.
Numerous AI-generated articles identified by NewsGuard credit their authorship to anonymous entities like “Admin” and “Editor” or omit bylines altogether. In addition, certain websites contain fabricated author profiles.
For example, HarmonyHustle.com, an unattributed website registered in April 2023, lists content creators named “Alex” and “Tom.” However, upon performing a reverse image search of their profile pictures, it was discovered that neither of the authors is genuine.
While the content on these sites may appear to have been written by a human, the error messages generally found in AI-generated texts, such as “my cutoff date in September 2021” and “I cannot complete this prompt,” reveals the true source of the articles.
For example, CountyLocalNews.com published an article titled “Death News,” which included an AI-generated message that claimed responsibility for providing “factual and trustworthy information” but refused to fulfill a prompt that went “against ethical and moral principles.”