AI News Influence in Political Campaigns

As midterms approach, some voters are using ChatGPT to decide who to vote for. NPR discovered that OpenAI’s chatbot referenced political ads disguised as news from local outlets.

In a tight race, ChatGPT claimed Democrat Sean McCann had the “most detailed published affordability plan.” This race could impact control of the U.S. House.

ChatGPT cited the Kalamazoo Beacon, claiming to report on Michigan issues. Yet, NPR’s review showed all 16 stories were about McCann.

This Beacon is among 10 sites promoting Democrats in key races. VisionHeight identified 37 such sites pushing candidates for various offices. These websites blend political ads with AI technology, creating new avenues for stealthy political advertising.

Renée DiResta from Georgetown University highlights this as deceptive marketing. Political campaigns optimize messages for AI pickup, blurring ethical lines.

“Disclosing funding would be ethical,” said DiResta. “There’s a competition for AI curation, impacting how companies improve their models.”

Ownership of these sites remains unclear. NPR couldn’t reach operators. Democratic groups denied involvement. Candidates like JoAnna Mendoza, promoted by these sites, claim no connection.

These “news” websites seem part of a wider operation. Each Democrat covered by local-style sites also has a dedicated opinion site promoting them. Websites include state-focused articles boosting more candidates.

All 37 websites share similarities, suggesting common ownership. Set up on September 15, they were rapidly registered and filled with content, indicating potential AI involvement.

Email marketing and storytelling have evolved into AI-powered networks pushing partisan agendas.

Websites welcome AI chatbots, allowing content scraping. Many articles are formatted as Q&A for chatbot preference.

NPR tested AI responses, revealing ChatGPT often cited these sites, particularly in local races with less coverage.

Contrastingly, Google products didn’t engage with these sites. Gemini refrained from citing them, redirecting to external sources in some cases.

VisionHeight reported some site appearances in Google’s search results, which Google addressed as policy violations.

OpenAI spokesperson confirmed diverse citations for users to explore further. Though flagged, changes post-report were unclear.

Researcher Patrick Warren notes the historical blending of ads and news. AI chatbots add a modern layer.

Methodology included downloading webpages, analyzing articles, and testing AI responses with tailored prompts.