Slopaganda: What the AI Ads of 2026 Tell Us About Elections
AI-made political advertising has arrived at scale in the US midterms, mostly undisclosed and often more absurd than deceptive. Brazil and Hungary suggest the real risks, and the real defences, lie somewhere other than the ads themselves.
A month before the US midterm elections of 3 November, artificial intelligence has become an ordinary tool of political advertising. The Wesleyan Media Project counts at least 164 ads produced or enhanced by AI in this cycle, NPR reported in mid-September. In Texas, Ken Paxton’s campaign ran an AI video of a Democratic rival in the Senate race. In Ohio, an AI ad depicted Sherrod Brown. Researchers quoted in the coverage describe much of this material not as sophisticated deception but as “slopaganda”: cheap, synthetic, often strange content produced in volume.
A second tally points the same way. An academic tracking project, writing in The Conversation, reports having tracked about $80m of spending on almost 170 AI-generated political ads this year, most of them without disclosure, according to the article’s headline and summary. The figures have not yet been independently confirmed in full, and the project’s methods will matter. But both counts suggest that AI advertising has moved from novelty to routine, and that the rules meant to govern it are not keeping pace.
Rules without a referee
The regulatory picture is fragmented. The National Conference of State Legislatures counts 31 states with laws on deepfakes or disclosure in political communications. At federal level the Federal Election Commission has deadlocked on regulating deepfake ads, Bloomberg Law reports. That leaves election officials, broadcasters and platforms to carry most of the burden in the final weeks of the campaign.
If the tracking project’s finding that most AI ads carry no disclosure is confirmed, it would mean state laws are either not being complied with or not being enforced. Disclosure rules depend on someone checking. A patchwork of state requirements, applied to ads that circulate nationally online, is hard for any single regulator to police.
Brazil: strict rules, persistent copies
Brazil offers the most demanding test of the alternative. Ahead of its first-round vote on 4 October, in which some 159m people were eligible to vote, the electoral court (TSE) required labels on AI campaign material, banned AI impersonation of candidates and imposed a 72-hour blackout on AI political content before the vote. These are among the strictest AI election rules anywhere.
They did not stop “Dona Maria”, an AI-generated avatar that criticised President Lula and drew millions of views, the Christian Science Monitor reported. The AI in Elections Observatory logged 29 cases. The avatar’s main accounts were removed in May, but copies persisted. The Monitor’s reporting captured the underlying problem in a phrase: “regulating individual users is borderline impossible.”
Rules aimed at campaigns can be enforced; rules aimed at everyone who can open an AI app mostly cannot.
Hungary: saturation without victory
Hungary provides a different lesson. In its election this spring, the governing Fidesz party used AI-generated videos and images at scale, targeting Péter Magyar and campaigning on Ukraine and the European Union, according to the European Digital Media Observatory. In polling cited by EDMO, 90% of respondents found AI campaign content “entirely unacceptable”, and 52% said they could not tell whether a deepfake was real.
Fidesz lost. EDMO’s analysis argues that political competition and independent media blunted the effect of AI content. That does not prove AI advertising is harmless; voters’ inability to identify deepfakes is itself a warning. But it is evidence against the simple narrative that deepfakes win elections.
A shifting American electorate
AI is also becoming a campaign issue in its own right. A Pew Research Center survey published on 16 September found that, for the first time, Democrats are more worried than Republicans about AI: 56% of Democrats said they were “more concerned than excited”, against 49% of Republicans. Among liberal Democrats, concern rose from 45% in 2023 to 63%. About half of all US adults now say they are more concerned than excited, up from 37% in 2021.
That realignment could change the politics of AI regulation, including the federal debate over overriding state laws. It may also mean that AI-made ads carry reputational risk for the campaigns that use them, particularly with voters already sceptical of the technology.
What the evidence suggests
Three conclusions emerge, each provisional. First, volume is not the same as impact; “slopaganda” may saturate feeds without changing many minds. Second, disclosure laws without enforcement capacity produce little disclosure. Third, the strongest defences appear to be structural rather than technical: competitive elections, independent journalism and rules targeted at those who pay for and distribute political ads, rather than at every individual user.
What to watch
- Publication of the full methodology and disclosure rate behind the $80m tracking figure.
- Whether the FEC breaks its deadlock before 3 November.
- Enforcement actions under any of the 31 state laws.
- Brazilian electoral authorities’ assessment of AI content during the 4 October first round.
Sources
- NPR — AI political ads in the 2026 midterms
- The Japan Times (NYT syndication) — AI political ads in the US midterms
- Bloomberg Law — Regulate deepfake ads, election agency told again after deadlock
- The Conversation — We’ve tracked $80m of spending on almost 170 AI-generated political ads this year
- Christian Science Monitor — AI in Brazil’s election
- EDMO — Hungary’s 2026 election: AI-driven post-reality campaigning and its limits
- Pew Research Center — Democrats are now more worried than Republicans about AI
Discussion
No comments yet. Start the conversation.
