This election has shown it’s time for us to get specific and critical on AI. The rules on AI-made political ads come from a mix of regulators, tech platforms and the parties’ own promises.
The Electoral Act requires every election ad to carry a promoter statement naming who’s behind it. It doesn’t require anyone to say an image was made with AI. “The Electoral Commission doesn’t regulate the content of election advertising,” the Commission’s social media guidance says.
The one content rule with criminal weight is section 199A of the Act. It makes it a corrupt practice to publish a statement you know is false in a material particular to influence a vote, but only from two days before polling day until the poll closes.
Andrew Lensen, a senior lecturer in AI at Victoria University of Wellington, and Andrew Geddis, a law professor at the University of Otago, say that safeguard is weaker now that voting runs over 12 days. They want AI disclosure on election ads and the false-statement rule extended across advance voting.
No Bill before Parliament covers election deepfakes. ACT MP Laura McClure’s Deepfake Digital Harm and Exploitation Bill, at select committee since May 2026, deals with sexual images made without consent.
What the parties have done
In 2023, National used AI-generated people in attack ads, including masked figures appearing to ransack a store above a line on crime. Christopher Luxon, then opposition leader, said an AI image was no different from a stock photo of an actor and saw no reason to disclose it.
A National spokesperson told 1News this year that the party sometimes uses AI content on social media, tries to label it, and doesn’t condone depicting real people in misleading ways. Newsroom reported in August that National’s 18billion.co.nz site attacking Labour’s tax plans used AI-generated imagery with no AI disclosure.
ACT was criticised in 2025 for an Instagram post using an AI-generated stock image of a Māori couple. ACT whip Todd Stephenson said AI imagery wasn’t inherently misleading and the party would never use AI to impersonate a real person. University of Waikato anthropology lecturer Bronwyn Isaacs says parties should be legally required to acknowledge AI in their ads.
Labour and the Greens say they don’t use AI to generate campaign images. NZ First says it will label any AI visuals, and Te Pāti Māori says its MPs are often the subjects of fake AI images.
Does a label fix it?
Simon Clark and Stephan Lewandowsky of the University of Bristol ran three experiments with 175, 275 and 223 UK participants, published in Communications Psychology in January 2026. People watched a deepfake video of someone confessing to a crime or moral wrong. Most still relied on it when judging guilt, even when warned beforehand that it was fake, although the authors say task framing may explain part of the effect.
Others put the weight on enforcement and free speech. ACT campaign manager Callum Purves says bad actors who spread AI misinformation should be held to account by the Electoral Commission and the media. The Advertising Standards Authority, which covers paid ads only, weighs freedom of expression under the Bill of Rights Act and gives political opinion more room than factual claims.
South Australia’s laws, in force since October 2025, ban AI deepfake election ads showing a person doing something they didn’t do without their permission, and any ad using AI must be labelled.
For business, the platforms already set disclosure rules. Meta’s policy requires political advertisers to disclose realistic AI-made images, video or audio of people or events, and since June 2026 it has used automated detection to add an “AI Info” label.
Google makes verified New Zealand election advertisers declare synthetic or altered content.
The ASA’s August 2026 AI guidance note covers all advertising. It says disclosure should be considered when AI creates a realistic person who doesn’t exist or imitates a real person, and that a label alone won’t make a misleading ad compliant.
The Electoral Commission asks voters to check who’s behind a post before trusting it, and to report missing promoter statements or misleading information about when, where or how to vote.
SPONSORED
