Ahead of the midterm elections, AI chatbots are directing voters to a network of websites that mimic local news outlets but function primarily as political advertisements for Democratic candidates. An analysis by NPR found that when users asked ChatGPT about candidates in competitive House races, the chatbot frequently cited a group of 37 websites, 10 of which are styled as local news sources promoting one Democrat each. The phenomenon highlights a new frontier in political campaigning where optimizing content for AI retrieval may influence voter perceptions in races that could determine control of Congress.
The websites, identified by digital security firm VisionHeight, were all registered on September 15, with all but one created within an eight-minute window. NPR’s review showed that ChatGPT cited articles from this network in 19 out of 37 queries, or roughly half the time, often alongside legitimate sources like Ballotpedia. In 17 of those instances, the chatbot specifically cited one of the outlets pretending to be a local news site, such as the Kalamazoo Beacon, which published 16 stories exclusively about Democratic candidate Sean McCann in Michigan’s 4th Congressional District.
What the Left Is Saying
Democratic operatives and candidates have largely distanced themselves from the websites, suggesting the operation may be independent of official party infrastructure. The Democratic Congressional Campaign Committee denied involvement in the creation of the sites when contacted by NPR. JoAnna Mendoza, a Democratic candidate in Arizona promoted by one of the sites, stated she was not behind the websites and did not know who operated them.
While some progressive analysts acknowledge the need for transparency, the perspective from the left often focuses on the broader context of information ecosystems. Renée DiResta, an associate research professor at Georgetown University who studies online influence campaigns, noted that opposition research entities are also trying to ensure negative information about candidates is surfaced by AI. She argued that while the current tactic crosses a line into deception by not disclosing funding or affiliation, it reflects a growing competition for curation in the AI age.
What the Right Is Saying
Conservative commentators and researchers view the network as a coordinated effort to manipulate algorithmic visibility, potentially giving Democratic candidates an unfair advantage in low-information races. The fact that the sites are openly partisan, with names like 'Public Safety Democrats' and 'Democratic Women in the House,' has drawn scrutiny from those who argue that AI platforms should better filter out such content. VisionHeight, the firm that identified the network, pointed to indicators suggesting a common operator, including the batch registration of domains and the use of AI to generate a high volume of content in a short period.
Critics on the right argue that the lack of transparency violates the spirit of journalistic neutrality. Patrick Warren, a professor at Clemson University's Media Forensics Hub, noted that while thinly veiled advertising is not new, its integration into AI chatbot responses represents a significant shift. He suggested that users reading AI answers without evaluating underlying sources are vulnerable to these tactics, describing it as 'nothing new under the sun' but amplified by new technology.
What the Numbers Show
NPR tested 37 specific questions across ChatGPT, Gemini, and Google Search. ChatGPT cited the network in 19 responses, or about half the time, with 17 of those citations coming from the 'local news' style sites. In contrast, Google's Gemini did not cite any links from the group in any of the responses tested, and Google Search did not surface the websites on the first page of results during the test period. However, VisionHeight noted that in early October, at least one site ranked higher than Ballotpedia on Google's first page.
The network consists of 37 websites in total. Ten are styled as local news outlets, each promoting a single Democratic House candidate. Another 12 focus on state-level races, and five are roundup sites with partisan names. All websites in the network included an 'llms.txt' file, a plain-text list of articles designed to help AI crawlers index content, a feature most standard news organizations do not currently employ. The sites explicitly allow AI crawlers from major tech companies to scrape their content, unlike many traditional news outlets that block such access to protect revenue.
The Bottom Line
The integration of partisan content into AI-generated answers raises questions about the neutrality of emerging search tools. OpenAI spokesperson Nate Evans stated that the company takes action when issues are brought to their attention, but NPR found that the Kalamazoo Beacon was still cited as a source for Sean McCann in follow-up tests. Google spokesperson Davis Thompson said the company detected a policy violation and took 'swift action' to block spam and abuse in its generative AI features.
As more voters turn to chatbots for political information, especially in down-ballot races where traditional media coverage is sparse—a phenomenon known as a 'data void'—the ability to influence AI outputs may become a critical campaign strategy. The ethical implications of undisclosed political advertising within AI responses remain a subject of debate among researchers and platform operators, with calls for greater disclosure of funding and editorial affiliation growing louder as the election approaches.