AI Chatbot Cites Websites Made to Look Like Local News in House Races
ChatGPT, an artificial intelligence chatbot, cited websites that present political ads as local news when asked about candidates in competitive U.S. House races, NPR reported.
In one example, ChatGPT linked to a website called the Kalamazoo Beacon as a source for its answer on Michigan's 4th Congressional District. When asked which House candidates in that district care about affordability, the chatbot ranked Democrat Sean McCann first and described his plan as the "Most detailed published affordability plan."
The Kalamazoo Beacon describes itself in its site code as "Reporting on Kalamazoo, southwest Michigan, and the issues shaping lakeshore communities." NPR found that the site has published 16 stories, all about McCann, his policy positions and his record as a state legislator. The site's code also labels its articles as "news articles."
A digital security firm, VisionHeight, identified a larger group of 37 websites. Most promote Democratic candidates for federal, state and local offices, and the group includes nonpartisan posts such as judges. Ten of the "local news" websites each promote one Democrat in a competitive House race.
NPR reported that Google's products avoid promoting these websites, while ChatGPT's responses pick them up. NPR said it was unable to find or reach the operators of the sites, and their ownership remains unclear.
Arizona candidate JoAnna Mendoza said she was not behind the websites promoting her and does not know who operates them. The Democratic Congressional Campaign Committee, the party's main committee for House races, denied involvement in the websites in a phone call with NPR.
Earlier this fall, RFI reported that partisan news sites it described as "pseudo-journalism" were targeting the U.S. midterm elections. The report said so-called "pink slime" sites, which imitate local news, are secretly backed by partisan groups on both sides of the political divide.
