The rapid development of artificial intelligence and especially large language models has now reached a stage, where generative AI is both powerful and accessible enough to be used at scale in election campaigns, at least in principle. In this paper, I outline some of the potential legal uses of AI by election campaigns: Generative AI already cuts costs for campaigns by assisting them at drafting communication such as fundraising emails or text messages. In the near future, the use of generative AI can potentially transform the way campaigns interact with voters by allowing them to conduct AI-to-voter conversations in many languages, and potentially at scale. While campaigns have personalised and tailored digital communication before the advent of mass-accessible generative AI-tools, the ability for LLMs to engage dynamically with information and arguments provided by voters is new and, potentially, transformative. The question is whether AI-to-voter communication via peer-to-peer messaging is scalable, and the answer depends on how access to personal data is regulated, which varies between countries and regions. If campaigns are legally able to access personal contact data, AI-to-voter communication is potentially disruptive, and could impact the political playing field in the mid-term.