If you have spent any time scrolling through social media in recent months, you have probably noticed a strange shift. Posts that look like real news, real photos, or genuine personal recommendations are everywhere, but something is slightly off. The images are too smooth, the headlines are oddly generic, the videos do not quite add up. This is the age of “AI slop”—the term Macquarie Dictionary has chosen as its Word of the Year for 2025. It is the result of generative artificial intelligence being used at massive scale, often deliberately, to produce content designed to mislead, distract, or simply grab clicks. It now floods the internet’s busiest platforms, burying thoughtful reporting under an avalanche of cheap, algorithm-baiting material. In New Zealand, this is no longer just a nuisance. After the tragic landslide at Mount Maunganui in January, AI-generated fake images of the disaster were shared thousands of times, spreading misinformation at a moment when people desperately needed accurate information. That was a warning shot. Later this year, New Zealanders will go to the polls, and every sign suggests the problem is about to get much worse. Political parties are already toying with these tools, and their early experiments show how tempting they are. For years, experts have warned about the risks of AI-driven misinformation. Now the use is accelerating, and the legal framework meant to govern it is struggling, perhaps failing, to keep up. We are entering an election season unlike anything we have seen before—not because candidates are more dishonest or voters less engaged, but because the tools of deception have become infinitely cheaper, faster, and harder to trace. The old rules of political communication are being rewritten, and no one has yet decided what the new ones should be. It is no longer possible to assume that a photo is proof, or that a video is a window onto something that actually happened. The result is a public sphere where genuine journalism and manufactured fantasy compete on equal footing, and where the effort required to create falsehood has dropped to almost nothing. That should worry everyone who cares about what is true.
At the centre of this concern is the use of AI in political campaigning. On Facebook, bogus news sites have already shared deepfake videos that appear to show New Zealand politicians meeting at Waitangi and making policy announcements, even though the scenes never happened. Some of these videos are crude, but others are unnervingly believable. More troubling still, political parties themselves have begun using AI to attack their opponents. The National Party, for example, has faced criticism for posting AI cartoon images of opposition leaders and for creating AI-generated attack advertisements during the 2023 campaign. Other parties are uncertain about how far they want to go, which means the electoral playing field may be dangerously uneven. One candidate might follow old norms while another floods the zone with machine-generated smears. Political attack ads have been around for as long as politics itself, but never before have they been so cheap and easy to produce. In the past, creating an effective smear campaign required money, imagination, and professional illustrators or media consultants. Now, with the right AI tools, almost anyone can generate a coordinated stream of false or misleading material in an afternoon, at virtually no cost. Even if political parties try to keep their hands clean, third-party lobby groups can do the dirty work for them, with little accountability. Foreign actors could also use AI to interfere in New Zealand’s electoral process, producing content designed to amplify division, suppress turnout, or sway a handful of votes in key electorates. In a contest as closely fought as the 2026 campaign is likely to be, a few thousand fraudulent posts could make the difference between winning and losing. And because these posts can be generated in countless variations and served up individually to different voters, tracing them back to their original source becomes almost impossible. This is not a hypothetical concern; the infrastructure for it already exists, and it is improving every day. It has become a cheap, infinitely scalable form of warfare, and the battlefield is the mind of the voter.
Beneath these practical concerns lies a much deeper problem: AI can fundamentally distort how voters perceive reality. Human beings are not good at distinguishing between genuine images and sophisticated fakes, especially when the content reinforces what we already want to believe. Research has shown that people are more likely to say someone is guilty of a crime when shown an AI-generated image of the person committing it—even after they have been told the image is fake. The image plants a seed of doubt that rational knowledge cannot fully remove. Now imagine that dynamic applied to politics. A voter sees a fabricated video of a candidate saying something outrageous. Even if they later learn it was AI, the emotional impression lingers. At a time when trust in politicians is already dangerously low, this kind of manipulation can deepen cynicism and push people away from civic life. If voters feel that everything they see in a campaign is fake, why bother participating? Another troubling development is hyper-personalised campaigning. By profiling voters through their social media activity, AI tools can generate custom messages tailored to each person’s specific fears, hopes, or frustrations. A candidate might send one message to a first-time buyer worried about housing, and another to a farmer frustrated with environmental rules—all automatically, all in seconds. In theory, this sounds like efficient communication. In practice, it means campaigns can find and exploit each voter’s psychological vulnerabilities without any public scrutiny. The public never sees the full picture; each voter sees only one piece of a hidden mosaic. This is not just another form of free speech. It is a transformation of political communication, and it is happening with almost no transparency or consent. We are not rational consumers of information; we are emotional, and AI knows how to push emotional buttons. The erosion of trust is not merely a side effect of this technology—it is often the goal. When people no longer know what is true, they stop trying to know anything at all, and that apathy is exactly what manipulators rely on. Without a shared basis of fact, democracy becomes impossible, because citizens can no longer debate the issues that matter. They just shout past each other in separate realities.
So what are the rules protecting New Zealand elections? In principle, the country has a solid foundation. The law regulates “election advertisements” in any media, and the definition is extremely broad: it covers any message that directly or indirectly promotes or opposes a party or candidate. There are also safeguards. Every election advertisement must include a “promoter’s statement” identifying who is responsible for it. Spending caps apply to the production and publication of ads during the three months before election day. And if an advertisement promotes a party or candidate by name, it must have that party’s or candidate’s written permission. These rules were designed to ensure accountability and prevent a few wealthy players from overwhelming the public debate. But the gap is in the content. There is nothing in the law that requires a candidate to disclose whether artificial intelligence was used to create an advertisement. There is no general prohibition on publishing false or misleading election advertisements. In New Zealand, you can legally lie in a campaign ad, as long as you do not trip over one of the narrow exceptions. The law focuses on who paid for an ad, not on whether the ad is true. One of the few exceptions is the rule that, in the final three days before the election period ends, it is an offence to publish a statement you know to be false if your aim is to influence how people vote. That sounds reassuring, but the rule is weaker than it appears. Voting in New Zealand now takes place over 12 days, not one. So a false statement published eight days before election day—when a significant portion of voters have already cast their ballots—may escape the rule entirely. The legal timing no longer matches the real-world rhythm of an election. Meanwhile, the fact that a deepfake does not always look like an advertisement makes the promoter statement requirement even less effective. If a deceptive image is shared as a news article, a meme, or a leaked video, most voters will never know who was behind it. As a result, the existing rules give voters little real protection when the most dangerous content is precisely the material that is easiest to disguise.
There is another old law that could theoretically help. The ban on “undue influence” over voters was designed to stop people from using force, threats, or fraud to prevent others from casting their votes freely. In an AI age, you could imagine it being used against deliberately misleading content intended to suppress turnout or make certain voters believe they have to vote on the wrong day. But the law dates back to the nineteenth century, has rarely been used, and was never designed for digital disinformation. Relying on it would be a gamble. So what should be done? At the very least, New Zealand should introduce rules that require election advertisements to disclose the use of AI, so voters can make an informed decision. This is not censorship; it is transparency. Another simple but effective measure would be to extend the “no deliberate lies to influence voters” rule to cover the entire advance voting period. If lying in the final three days is illegal, there is no logical reason to allow it earlier, when many voters are already casting their ballots. New Zealand can also learn from other countries that are being forced to respond. The European Union and many US states have recently passed laws specifically addressing deepfakes in political campaigns. Closer to home, the Australian Electoral Commission has created a “disinformation register” to combat false claims about how the electoral process works. That is a useful step, though it is limited: it deals with the mechanics of voting—where and when to vote—rather than false statements about parties or candidates. New Zealand has an opportunity to go further, not only by borrowing what works abroad but by tailoring reforms to its own political culture. The goal should be to protect the honesty of public debate while still allowing creative expression and legitimate campaign speech. It will not be easy, but the alternative is to surrender one of democracy’s most basic safeguards: the ability of citizens to believe their own eyes. The law should never make it impossible to identify who is trying to manipulate an election.
The bigger picture is difficult to ignore. Around the world, democracy is under strain, and public confidence in democratic institutions is falling. In New Zealand, that decline is not as steep as in some countries, but it is real. Protecting the integrity of elections should be a top priority for anyone who cares about the country’s future. Free and fair elections are not simply about counting votes accurately; they depend on transparency, trust, and an electorate that can make decisions based on good information. Artificial intelligence is already testing those foundations, and the test will only become more intense in the months ahead. Updating the rules will not solve every problem. Technology is moving fast, and any legislation will quickly find itself catching up. But doing nothing is not an option. Without new rules, the 2026 election could become a laboratory for AI-driven manipulation, with consequences that extend far beyond who wins or loses. The damage to trust, social cohesion, and the simple idea that citizens can share a common factual reality may take a generation to repair. The good news is that New Zealand still has time to act before the campaign reaches full speed. It is not a matter of banning AI; generative tools have useful, creative applications in politics and elsewhere. The aim should be disclosure, accountability, and fairness. New Zealand can choose to be a leader on this issue, to prove that a small democratic country can adapt to the hardest challenge of the digital age. Doing so requires courage and political will, but the alternative—watching the next election drown in a sea of indistinguishable real and fake, true and false—is unthinkable. The tools have changed, but the values that should guide politics have not. It is time for the rules to catch up. If they do, New Zealand can still navigate the AI age with confidence. If they do not, the consequences will be measured not only in election results, but in the slow erosion of what we expect from one another as citizens.