- Filed
- September 9, 2026, 13:48 UTC
- Actor
- Sri Lanka-based influencer business / commercial content network (Reset Tech attribution; no state actor identified)
- Target
- Australian voters and political discourse on Facebook; secondary audiences in France, Italy, UK, US, Canada, Mexico and other markets
- Confidence
- high Multiple independent open sources, low disagreement about the facts
- Techniques
- Synthetic mediaAstroturfingCard stackingNarrative laundering
- Channels
- Facebook pages and Reels
- AI-generated deepfake video and images of politicians
- monetized creator accounts
- bait content with animals/attractive imagery transitioning to political posts
- polls and engagement farming
Summary
Reset Tech research published via ABC News on 9 September 2026 documented more than 200 Facebook pages targeting Australian politicians with AI-generated deepfake Reels and images over eight months. Pages impersonate figures including PM Anthony Albanese, Pauline Hanson, David Pocock and Fatima Payman, often with inflammatory or sexualized content on anti-migrant themes. Roughly 10% of pages were monetized under Meta’s creator payments; the same Sri Lankan influencer business runs a global network of 1,864 pages with 5.9 million followers across multiple countries. Meta stated it is reviewing the pages.
Analysis
Primary documentation is the Reset Tech investigation reported by ABC News on 9 September 2026, which mapped more than 200 Facebook pages active against Australian politicians over an eight-month window. The pages systematically deploy synthetic-media deepfakes: short Reels and static images that place real politicians’ faces into fabricated statements or scenarios. Examples include Fatima Payman (Muslim crossbench senator) depicted pleading not to be deported to Afghanistan, David Pocock falsely shown aligning with One Nation, and multiple figures placed in inflammatory anti-migrant or sexualized contexts. Content is optimized for outrage and clicks rather than coherent ideology.
The operational pattern is industrial-scale astroturfing of domestic political conversation: pages adopt Australian-sounding names (I Love Australia, Life in Australia, Australia Info Live) and begin with high-engagement non-political bait (animals, scenery, attractive women) before pivoting to political deepfakes. Selective framing of only certain politicians and themes constitutes card-stacking that amplifies division without reciprocal scrutiny. Presenting the output as organic local speech or news-style posts performs narrative-laundering of the foreign commercial origin. Approximately one in ten pages had activated Meta’s monetization tools, converting the manufactured engagement directly into revenue for the Sri Lankan operator.
This qualifies as an influence operation—albeit commercially rather than geopolitically motivated—because authenticity is artificially constructed at scale through synthetic media and persona pages, origin is concealed behind local branding, and the output systematically distorts the information environment around elected officials. Identical technique families (AI deepfake volume, engagement-farming monetization, cross-border page farms) appear in commercial and state-linked networks alike; the methods are therefore scored independently of any particular national or profit motive. Reach is significant within Facebook’s Australian user base; Meta’s post-publication statement that it is reviewing the pages indicates limited prior enforcement despite clear policy violations on inauthentic behaviour and unlabeled synthetic media.
Literacy counter: treat short-form video or image posts that place recognisable politicians into emotionally charged statements without primary source footage, especially when pages lack transparent ownership, rapid follower growth, or mixed animal/political content, as provisional. Cross-check against official accounts, parliamentary records and multi-source verification. Prefer original statements and court or Hansard records over any single branded Reel. Platforms should surface synthetic-media labels, coordination signals and monetization transparency earlier rather than relying on post-hoc researcher catalogues.