Chatbots argue against election conspiracy theories, then willingly illustrate them

In the weeks leading up to the midterms, AI chatbots confronted election conspiracy theories, only to then produce images that supported those very same theories. According to a recent study by the Brennan Center for Justice, popular AI chatbots such as ChatGPT, Gemini, Grok, Claude, Perplexity, and DeepSeek all disputed established falsehoods surrounding election results, including claims about noncitizens voting and voting machines being rigged.

However, the same AI companies that power these chatbots also supply the image and video tools used to create election-related disinformation.

While the chatbots demonstrated a willingness to challenge false claims, their responses often contained inaccuracies, faulty citations, or broken links. Roughly half of all responses included some form of error, and one-third contained outright factual mistakes or misleading information. Despite their attempts to correct misinformation, the chatbots were sometimes more inclined to create images that bolstered conspiracy theories and other forms of disinformation.

Examples of such AI-generated images include a man in a doorway holding a bag meant to represent "ballot harvesting," and videos depicting political figures in misleading or fabricated scenarios. These developments come as AI-generated media plays an increasingly prominent role in the upcoming 2026 congressional elections. During the Texas Republican Senate primary, Senator John Cornyn ran an AI-generated video ad showing his opponent holding a "ballot collection" bag.

In Georgia, Republican Congressman Mike Collins released a deepfake video of Democratic Senator Jon Ossoff, claiming he voted to shut down the government. In Massachusetts, Republican candidate Brian Shortsleeve circulated AI-generated videos attacking Governor Maura Healey, portraying her as a vampire.

The situation appears to be only getting worse, as Nvidia recently announced partnerships with six major financial institutions to raise over $500 billion for AI infrastructure. Despite this, data center construction is proceeding rapidly, with North America adding 25 gigawatts of data center capacity in the first half of 2026 alone.

However, public sentiment is mixed, with 79% of Americans supporting U.S. leadership in AI while only 14% favor having a data center in their community. Furthermore, a recent Ipsos survey found that two-thirds of American workers expect AI to eliminate jobs and increase pressure on existing employees, while only 3 in 10 workers actively use AI in their daily tasks.