LLMs Become an Election Variable

Ahead of the 2026 elections, US political circles believe that LLMs such as ChatGPT and Claude could influence voters' judgments about candidates, but they still lack a clear response strategy for how to handle this.

Political strategists admit that it's hard to predict what an LLM will answer, and that methods for changing those responses are still in their early stages. The Democratic-aligned American Bridge 21st Century is experimenting with AEO (answer-engine optimization), which aims to shift LLM responses in a favorable direction similar to search engine optimization (SEO), but publicly known efforts remain limited.

According to the article, when NOTUS posed political questions, ChatGPT and Claude summarized candidates' criminal records, policies, and scandals, offering relatively direct judgments. For example, regarding Roy Cooper in the North Carolina Senate race, both criminal issues and his moderate image were mentioned together, and among potential 2028 presidential contenders, figures with heavier media exposure were given more weight.

Some campaigns and organizations have already begun responding. Trends are emerging where tools like Caucus AI are being built to track answers from Gemini, Grok, and ChatGPT state by state, and preparations are being made to gain better visibility in sources such as candidate websites, organized policy documents, Wikipedia, and YouTube/X. Still, the core point of the entire article is that while LLMs are becoming a new gateway for political information, there is not yet a standard strategy to systematically control this.