AI Citations: 37 Websites Registered in Minutes Promote Democratic Candidates
A network of sites masquerading as local news outlets is being cited by ChatGPT in competitive House races, raising questions about transparency in AI-generated political information.
By Megan Parker ·

Voters increasingly rely on artificial intelligence to research candidates, a habit that has exposed a network of websites designed to look like local news but function as partisan advertising. Ahead of the November midterms, a review found that ChatGPT frequently cites these sites when users ask about specific congressional races. The phenomenon creates a new channel for political messaging that blurs the line between journalism and advocacy.
In Michigan’s 4th Congressional District, a competitive race for control of the U.S. House, ChatGPT identified Democrat Sean McCann as the candidate with the "Most detailed published affordability plan." The chatbot linked this claim to the Kalamazoo Beacon, a site that describes itself in its source code as reporting on southwest Michigan. However, the outlet has published only 16 stories, all of which focus exclusively on McCann, his policy positions, and his record as a state legislator.
The Kalamazoo Beacon is one of 10 sites that promote a single Democrat in a competitive House race. These 10 outlets are part of a larger group of 37 websites identified by the digital security firm VisionHeight. The broader network promotes numerous candidates for federal, state, and local offices, including nonpartisan posts like judgeships. While most of the sites are openly partisan, they link to sources such as congressional voting records and mainstream news articles to support their claims.
Digital forensics suggest a single operator controls the entire network. Registry and digital certificate data indicate that all 37 websites were set up on September 15. All but one were registered within an eight-minute period. Hundreds of articles appeared in batches during September and October, a volume that VisionHeight’s report suggests may indicate the use of AI to generate the material. The sites were refined in October, with their descriptions becoming more specific to their purported geographic areas.
The websites are engineered to be easily consumed by AI models. All 37 sites explicitly allow a long list of AI crawlers to scrape their content. They also include a file called llms.txt, which lists articles in plain text, a feature not common to most websites. Alisha Ukani, a postdoctoral researcher at the University of Washington, noted that sites with these files tend to be technology or business focused. The network’s articles are often formatted as questions and answers, a structure that chatbots prefer when citing sources.
To measure the network’s influence, a review tested 37 questions on ChatGPT, Gemini, and Google search. ChatGPT cited at least one article from the network in 19 of the responses. In 17 of those instances, the citation came from one of the sites posing as local news. Gemini did not cite any links from the group in its responses. Google search also did not surface the sites in its first page results, though its AI Overview feature did not mention them either. In early October, at least one site appeared on Google’s first page, ranking higher than Ballotpedia. Google’s Davis Thompson said the company detected a policy violation and took "swift action."
It remains unclear who operates the websites. The Democratic Congressional Campaign Committee denied involvement in a phone call. One candidate promoted by the network, JoAnna Mendoza of Arizona, said she was not behind the sites and did not know who was. Other promoted candidates, including McCann, and super PACs such as House Majority PAC did not respond to requests for comment. Renée DiResta, an associate research professor at Georgetown University, said that disclosing funding would be the ethical course of action. She noted that the situation creates a "competition for curation" as companies try to refine their models.
OpenAI spokesperson Nate Evans stated that ChatGPT’s search tools cite a range of sources to help users make their own judgments. He added that the company takes action to improve model responses when issues are brought to its attention. Since the sites were flagged, the Kalamazoo Beacon was still cited in subsequent queries. In some tests, McCann no longer received the distinction of having the most detailed plan, while in others, he did. Patrick Warren, a professor at Clemson University, described the tactic as a modern version of ad flyers that look like newspapers, noting that there is nothing new under the sun.
Source: WCBE






