Voters are increasingly turning to AI chatbots to inform them about political issues (Four Corners: Mark Hiney). When they ask models such as Claude or ChatGPT which parties stand for different policies, the responses are not always balanced.

A study by digital‑marketing agency Topham Guerin examined 200 AI‑generated answers that compared Victorian parties on key issues. Labor appeared in every answer, the Liberals and Nationals in 199 each, the Greens in 178, but One Nation was mentioned only 43 times.

Across the 155 recorded comparison answers, only four of the seven AI systems actually searched for One Nation. This shows that many tools rely on a narrow set of trusted sources, so parties with a smaller online footprint are less likely to be included.

Andrew Hirst, former Liberal Party director and now chief executive of Topham Guerin, explains: "Voters and consumers are increasingly getting their information from AI tools, so political parties and media outlets are being judged on how well a machine can understand them." He adds, "These tools aren't making things up from scratch. They draw heavily on a small number of sources they trust. Whoever explains something most clearly online ends up explaining it to everyone who asks an AI."

Ben Guerin, co‑founder of Topham Guerin, says: "This isn't bias in the way people usually mean it. If a party doesn't set out its policies clearly on its own site, and what the media and others say about them is thin, the AI has less to work with, so it says less. Being absent online means being absent from the answer chatbots give."

The research is a snapshot of September 22‑23, using developer interfaces of AI companies. It does not show that anyone changed their vote as a result of the AI answers.