New report reveals rapid growth in AI-driven political advertising

A fresh industry analysis has documented how swiftly artificial intelligence has moved from the margins of campaign strategy into the daily toolkit of political communicators. The study, compiled by researchers tracking advertising patterns across several democracies, found that machine-generated content now accounts for a meaningful share of the paid messaging that voters encounter on social platforms, streaming services, and connected television. Researchers argue that the trajectory points toward a near-future election cycle in which a majority of digital campaign assets are produced, optimised, or personalised with little direct human involvement beyond the initial brief.

For Australian readers, the findings land at a moment when federal and state campaign teams are openly testing generative tools, while regulators in Canberra weigh how to keep pace. The data offers a baseline for understanding what voters in Sydney, Melbourne, Brisbane, and regional seats are likely to see in their feeds over the coming months, and what the major parties have already begun purchasing on their behalf.

The scale of the shift documented in the report

According to the report's authors, the volume of AI-assisted political advertising more than doubled across the sample of campaigns examined, with the steepest growth occurring in the eighteen months leading up to recent national votes. Micro-targeted video variations, synthetic voiceovers, and algorithmically refreshed copy are flagged as the three formats that grew fastest. Researchers noted that smaller parties with leaner media budgets appear to be adopting these tools at rates that rival, and in some cases outpace, the larger established parties.

The report distinguishes between campaigns that openly label AI involvement and those that do not. In the Australian portion of the case studies, researchers observed a handful of state-level campaigns experimenting with AI-generated script drafts and voice variants during recent state elections, though much of this work remained behind closed doors. The pattern mirrors what is happening in comparable federations, where the technology has outpaced formal disclosure norms.

Australian campaigns and the voter response

Inside the major party war rooms, the appetite for generative AI has shifted from curiosity to operational necessity. Senior strategists in Melbourne and Sydney-based firms have begun building internal prompt libraries that allow junior staff to produce dozens of ad variants from a single approved message. The goal is not novelty but volume: a single policy announcement can now be reworked for different suburbs, age cohorts, and language groups within minutes rather than days.

Polling commissioned alongside the report suggests that Australian voters are sceptical but not hostile. A majority of respondents in metropolitan Sydney and Melbourne indicated they would be less likely to trust a political ad if they knew it was generated by software, yet they expressed greater comfort with AI-assisted translations and accessibility enhancements. Regional respondents in the Hunter and the Pilbara were less concerned about the technology itself and more focused on whether the underlying claims were accurate.

Several campaign operators have spoken about the appeal of synthetic voice tools for replicating an MP's vocal tone across multiple languages, including Mandarin, Vietnamese, and Arabic, reflecting the linguistic diversity of seats in Sydney's west and Melbourne's inner suburbs. Others have leaned on image-generation software to produce location-specific visuals for outer-suburban and regional electorates without booking photographers for each appearance. The economic logic is straightforward: the marginal cost of an additional ad variant is now close to zero.

Deepfakes and the question of integrity

The report devotes significant attention to the darker edge of this expansion. Synthetic video and audio, once expensive to produce convincingly, can now be generated by anyone with a laptop and a subscription. Researchers documented a small but rising number of cases in which fabricated statements attributed to candidates circulated during election periods before fact-checkers intervened. The reputational damage, even when a fake is debunked, tends to outlast the correction cycle.

Australian audiences have already seen early examples. Ahead of the 2023 referendum, altered clips circulated on encrypted messaging channels and short-video apps, prompting the Australian Electoral Commission to issue public guidance on verifying political content. Analysts warn that the next federal contest will almost certainly feature more sophisticated fabrications, particularly in tight inner-Brisbane and Perth marginal seats where every percentage point carries weight.

The regulatory landscape in Canberra

Australian law currently treats digital political advertising under rules shaped by the Electoral Act and the Privacy Act 1988, but neither framework was drafted with generative AI in mind. The Australian Communications and Media Authority has limited powers over platform-side targeting, and the AEC focuses primarily on disclosure thresholds rather than content verification. A review of electoral communication laws was flagged in recent policy papers, though a clear timetable has not yet been set.

Civil-society groups including the Australia Institute and academic centres at the University of Melbourne and UNSW have called for mandatory watermarking of synthetic media used in paid political messaging. Industry submissions to the federal government suggest that voluntary codes may emerge first, with formal rules following once the next parliament sits. Until then, parties and their suppliers operate in a grey zone that the report's authors describe as among the most permissive in the democratic world.

Tool categories reshaping paid political messaging

Several categories of AI application are growing faster than others, according to the underlying data. The list below reflects the formats with the highest adoption rates across the campaigns studied:

• Automated copywriting that produces hundreds of headline and caption variants from a single policy brief.
• Synthetic voice and avatar generation for multilingual outreach and accessibility.
• Predictive micro-targeting models that score voter segments based on likelihood of persuasion.
• Real-time creative optimisation that swaps visuals and calls to action based on engagement signals.
• Synthetic focus-group simulation used to test messaging before paid distribution.

These tools are typically layered on top of existing advertising stacks, meaning that a single campaign in a marginal seat can now combine traditional polling data, third-party voter files, and AI-generated creative in a way that would have required a large agency team only two cycles ago. The result, researchers argue, is a structural advantage for well-funded operations that already possess the underlying data infrastructure.

Indicators worth tracking before the next federal vote

The closing chapters of the report sketch a set of indicators that researchers believe will define the integrity of upcoming elections. The list below summarises the signals worth tracking:

• Disclosure of AI involvement in paid political messaging by major parties and independents.
• Platform-level policies on synthetic media labelling and political advertiser verification.
• Coordination between the AEC, ACMA, and the Office of the Australian Information Commissioner on AI-related complaints.
• Public reporting by fact-checking organisations such as RMIT's ABC Fact Check and AFP FactCheck.
• Litigation outcomes from any high-profile deepfake incidents during the campaign period.
• Cross-border cooperation with election integrity bodies in New Zealand, the UK, and Canada.

Researchers argue that the absence of movement on even one of these indicators by the time the writs are issued would suggest that Australia is entering its next federal contest with weaker safeguards than comparable jurisdictions. The report recommends that an independent monitor be appointed to publish quarterly briefings during the official campaign period, modelled on arrangements already in place for election finance disclosures.

Where readers can follow the story

The full dataset behind the report is expected to be released through an academic publisher later this year, with a public dashboard tracking AI-assisted political advertising in real time. Australian readers interested in tracking how these tools are deployed locally can follow coverage from ABC News, The Guardian Australia, and the Sydney Morning Herald, alongside ongoing analysis from the Parliamentary Library. Independent watchdogs including the Australia Institute and the Centre for AI Safety will publish quarterly updates on synthetic media incidents as the campaign calendar firms up.

As the next federal writ approaches, the choices made by parties, platforms, and regulators will shape whether voters in Parramatta, Pearce, and Rankin see political messaging as a transparent conversation or an algorithmically optimised fog. The growth documented in this report makes clear that the technology is already here; subscribe to our daily digest to stay informed as new evidence emerges and the rules around AI-driven campaigning continue to take shape.