{
  "id": 10286031,
  "title": "Deepfakes are distorting this year’s Victorian election. We found out who’s most susceptible to them",
  "url": "https://urgent.news/2026/09/27/deepfakes-are-distorting-this-years-victorian-election-we-found-out",
  "topic": "world",
  "section": "World",
  "published": "2026-09-27T19:19:17.000Z",
  "source": {
    "name": "The Conversation AU",
    "slug": "the-conversation-au",
    "url": "https://theconversation.com/deepfakes-are-distorting-this-years-victorian-election-we-found-out-whos-most-susceptible-to-them-291443"
  },
  "original_language": "en",
  "account": "The Victorian election is still months away, yet voters are already encountering events that never occurred. These events, such as a machete attack at a petrol station and a woman giving birth on the roadside, are deepfakes produced by artificial intelligence (AI) and reportedly appearing as paid advertisements ahead of voters heading to the polls. Once considered a potential \"sleeper issue\" in Australian politics, deepfake political content is swiftly becoming a genuine electoral challenge.\n\nOur preliminary research focuses on political deepfakes and how people respond to them ahead of the Victorian state election. Our findings provide crucial insights into people's ability to detect deepfakes, which will aid voters in navigating the complex information ecosystem as campaigning intensifies.\n\nDeepfakes used to require specialized technical knowledge and software to create convincingly. However, they can now be produced by anyone using commercial and open-source AI tools, making them increasingly difficult to identify. We conducted a meta-analysis combining findings from 2018 to 2026 studies involving over 24,000 individuals. The results, yet to be published, reveal that deepfakes can be just as persuasive as authentic political content, including real footage and conventional misinformation.\n\nMoreover, deepfakes are linked to diminished trust in news and democratic institutions. This uncertainty can contribute to the \"liar's dividend,\" where individuals start doubting genuine content. In a test involving 411 Australians, participants viewed authentic political videos, low-quality deepfakes, and high-quality deepfakes. Preliminary results are concerning: those who consumed more news were better at detecting low-quality deepfakes but less accurate with sophisticated ones. Notably, 75% of participants could correctly identify deepfakes, but this still leaves about one in four people who struggle with this task.\n\nEven more troubling, nearly 29% incorrectly labeled genuine videos as fake, and another 18% were uncertain. This suggests that nearly half of participants could not confidently and accurately identify real political content. The danger lies not only in believing false information but also in becoming too ready to disbelieve genuine content.\n\nWho is better at spotting a deepfake? The answer is complex. Previous research indicates that political deepfake detection decreases with age, which could be particularly relevant in Australia due to the significant proportion of older voters. Political orientation may also play a role, with left-leaning individuals potentially more accurate at detecting political deepfakes. However, our preliminary experimental findings suggest differences only when the deepfake depicts a political figure aligned with the viewer's political orientation. This implies that susceptibility may depend not just on a person's political leanings but also on who is being depicted and their alignment with the viewer's beliefs.\n\nConfidence in detecting deepfakes is mixed. Some studies suggest that greater confidence can assist people in identifying deepfakes, while others find that confidence has little relationship with correct identification. Our latest findings add another layer: among moderate and higher news consumers, confidence becomes increasingly disconnected from accuracy. These participants may feel well-equipped to judge political content, but their confidence does not necessarily translate into accurate identification. Feeling confident that something is AI-generated does not mean it actually is. In fact, it may become a vulnerability, potentially leading voters to question even authentic evidence because it could have been generated by AI.\n\nSpotting a political deepfake is not straightforward, as there is no foolproof method. As AI advances, voters should focus more on verifying the source and supporting evidence rather than relying solely on appearance. First, avoid relying on visual or audio flaws, as these cues for detecting deepfakes are becoming less reliable. Second, trace the source of the content and assess the credibility of the source. Third, seek independent reporting or official sources to confirm the content. Fourth, question your confidence—feeling certain something is fake does not mean you are right. Finally, refrain from amplifying suspicious content. Before sharing, verify its authenticity.\n\nAs political deepfakes become more sophisticated, spotting visual imperfections will become increasingly challenging. Voters must adapt by becoming better at questioning what they see without automatically disbelieving it. In an era of increasingly convincing AI, healthy skepticism is essential.",
  "summary": "Voters need to become better at questioning what they see, without automatically disbelieving it.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}