Paragraph 1
Picture this: you’re scrolling through your feed late at night, half-watching the usual clutter of campaign ads and viral videos, when suddenly you stop cold. There’s a U.S. Senate candidate—someone you’ve never heard of before—sitting at a table, laughing at a set of cards that read “Death to America” and “Celebrate 9/11.” The room looks ordinary, the lighting feels authentic, the man’s expression seems genuinely caught on camera. Your mind races: did he really just do that? How could anyone running for office be so reckless, so unpatriotic, so vile? But here’s the catch: the video isn’t real. It’s a deepfake, a piece of hyper-realistic artificial intelligence-generated content designed to look exactly like reality while showing something that never happened. And it’s not an isolated prank or a random internet hoax. It’s part of a multimillion-dollar blitz of AI-generated political ads rolling out across television and social media in the run-up to the November 3 midterm elections. Researchers have a name for this strange new genre of political communication: “slopaganda.” It’s sloppy, synthetic, and often shockingly convincing, and it’s being weaponized by both sides of the aisle to smear rivals, energize base voters, and muddy the waters of public debate. The result is a political landscape where seeing is no longer believing, and where even the most outrageous video might be a manufactured lie. This is not some distant, futuristic warning. It is happening right now, in this election cycle, in your feed, on your television, and in the last few anxious seconds before you click “share” and cry out in disbelief at the latest political scandal that never was.
Paragraph 2
Take the case of Abdul El-Sayed, a Democratic candidate for Senate in Michigan, whose campaign has been targeted by one of the most striking and disturbing AI-generated ads yet. The spot opens with a voiceover that says, “His family values are wrong for Michigan,” while viewers see El-Sayed playing a board game with two people wearing traditional kaffiyehs, the checkered Arab scarves that are often associated with Palestinian identity and solidarity. The ad is lifelike, unsettlingly so. It features cards bearing incendiary statements like “Death to America” and “Celebrate 9/11,” and it even displays a fabricated family photograph of El-Sayed surrounded by four women wearing head-to-toe veils. The implication is unmistakable: that El-Sayed is somehow foreign, dangerous, and sympathetic to America’s enemies. The voiceover drives the point home: “Destruction of Israel and the West isn’t just a game for them. It’s a mission.” None of this is true. El-Sayed did not pose for that photo. He did not play that game. The entire scenario is a digital invention, crafted by a political action committee called Front Line Action, which is aligned with Republican Senator Ted Cruz. According to filings with the Federal Election Commission, the group spent more than $92,000 on advertising opposing El-Sayed. To its credit, or at least in compliance with some basic transparency measure, the ad carries a small disclaimer saying it was “created in whole or in part with the use of artificial intelligence.” But that disclaimer is tiny, easy to miss, and buried in the visual chaos of the spot. For most viewers, the image is what sticks: a Muslim candidate laughing alongside symbols of terror. The damage is done before anyone reads the fine print. And this is just one example. Across the country, similar ads are popping up in races both big and small, each one designed to exploit fears, trigger outrage, and turn perception into a weapon.
Paragraph 3
What makes this moment especially dangerous is that most AI-generated political ads do not carry any label at all. Researchers who track election advertising have been racing to identify these spots, and their findings paint a sobering picture. The Wesleyan Media Project, a nonpartisan group that monitors campaign spending and ad content, has so far identified 220 political ads nationwide that were generated at least in part using AI tools. That number alone is staggering, but the spending behind it is even more so. The group estimates that these ads cost more than $137 million to produce and distribute. To put that in perspective, in the entire 2024 election cycle, researchers identified less than $50,000 in spending on AI-generated ads. The growth is not incremental; it is explosive. And the partisan breakdown is striking: 81 percent of the AI-created ads were sponsored by Republican candidates or groups, 12 percent by Democrats, and the remaining 7 percent by independent candidates or groups. That imbalance suggests that, at least for now, the strategic embrace of AI-generated content is being driven largely by one side of the political spectrum, though the technology itself is available to everyone. Erika Franklin Fowler, co-director of the Wesleyan Media Project, put it bluntly: “AI is making its way into the 2026 midterms in a big way this year.” She also noted that only 27 percent of the ads carried any kind of AI disclaimer. That means nearly three-quarters of these synthetic political messages are appearing without any indication that they were created by machines. The group acknowledged that its count is almost certainly an undercount, because it is often impossible to determine with certainty whether AI tools were used in the making of an ad. Some tells exist, of course—oddly rendered fingers, unnatural eye movements, slightly warped backgrounds—but the technology is improving so quickly that even trained researchers are struggling to keep up. The result is an information environment in which voters cannot trust their own eyes, and where the old assumption that “there’s a video, so it must be real” no longer holds.
Paragraph 4
The regulatory landscape around these ads is a patchwork of incomplete rules and voluntary disclosures. There is no federal mandate requiring political ads to reveal their use of artificial intelligence. Some states have passed their own laws restricting AI-generated content or demanding transparency, but these laws differ widely, and they cannot easily cross state lines or control what appears on national platforms. In Congress, the issue has begun to attract attention, but concrete action remains slow. Democratic Senator Jon Ossoff of Georgia has been one of the most vocal critics of the trend. Last month, he spoke out after demanding the removal of a fabricated ad that portrayed him receiving beauty treatments at a salon, with a voiceover suggesting he was trying to cover up his “ugly record.” The ad, notably, carried no AI disclosure at all. It was sponsored by the Senate Leadership Fund, a political action committee that describes itself as “solely dedicated to protecting and expanding the Republican Senate majority.” Ossoff called the use of deepfakes in political advertising “deeply troubling” and warned that it represents “a slippery slope toward the destruction of truth in political debate.” His concern is not hypothetical. The London-based Institute for Strategic Dialogue has described the current midterms as the first U.S. election in which AI is built into every major social media platform. That means the tools to create, distribute, and amplify synthetic content are no longer confined to sophisticated hackers or well-funded studios; they are available to anyone with an internet connection and a willingness to deceive. The platforms themselves have been slow to respond. Some have introduced labeling systems, but enforcement is inconsistent, and many ads slip through without any flag. Voters are left to navigate a media landscape where the line between reality and fabrication is increasingly blurred, and where the very mechanisms that once helped people make informed decisions—photographs, videos, audio recordings—have become unreliable witnesses.
Paragraph 5
Public opinion, at least, seems to be firmly on the side of transparency and truth. A recent survey by the nonprofit AAPC Foundation found that 73 percent of Americans believe it is “unacceptable” for a political campaign to use AI to generate images or video in an ad. That is a surprisingly high number in a deeply polarized country, where almost nothing unites people across party lines. Even more striking, 95 percent of respondents said it was unacceptable to “create a video of an opponent saying something they never said.” That near-universal consensus suggests that voters themselves recognize the danger of these tools, even if the campaigns using them do not. The disconnect is profound. On one hand, Americans overwhelmingly reject the use of AI to fabricate political content. On the other hand, campaigns and political action committees keep spending millions of dollars to do exactly that, presumably because they believe it works. And perhaps it does, at least in the short term. A viral fake video can generate millions of views, spark angry reactions, and dominate news cycles for days. Even after it is debunked, the emotional impression may linger. The skeptical viewer might think, “Well, where there’s smoke, there’s fire,” and the false image becomes embedded in their memory alongside the correction. This is what researchers call the continued influence effect: even when people are told that information is false, they continue to rely on it when forming judgments. In a political context, that means a deepfake ad can accomplish its goal—casting doubt, stirring anger, rallying supporters—even if it is eventually exposed as a lie. The people who create these ads are not counting on fooling everyone forever. They only need to fool enough people long enough to make a difference in an election.
Paragraph 6
Yet some researchers are pushing back against the panic, arguing that the real-world impact of AI-generated political ads may be more limited than it appears. A recent research report from New York University’s Center on Tech Policy suggested that “political slopaganda is likely more effective at mobilizing voters or raising money than it is in impacting candidate selection.” In other words, these ads may be better at firing up people who already support a particular side, or at generating small-dollar donations fueled by outrage, than at actually changing the minds of undecided voters. The NYU report also drew an analogy to internet memes, noting that “like digital memes, there may be limits to its attentional staying power, especially as we become more and more inundated with AI slop.” That is a comforting thought, but not entirely reassuring. After all, memes can be incredibly influential, and the sheer volume of AI-generated content is growing so fast that it may be hard for viewers to process everything they see. The real danger may not be any single ad, but the cumulative effect of living in a world where every video, every photo, every audio clip is suspect. When everything can be faked, nothing can be trusted. And when nothing can be trusted, voters may retreat into their own echo chambers, believing only what they want to believe and dismissing all evidence as manufactured. That is a recipe for democratic decay. So while researchers debate whether these ads actually flip votes or simply reinforce preexisting beliefs, the deeper question remains: what happens to a society when truth becomes just another casualty of political competition? The answer is uncertain, but the trend is clear. AI-generated political advertising is here to stay, and without stronger safeguards, more transparent labeling, and a public that is both skeptical and informed, the integrity of elections will continue to erode. The candidate laughing at cards that say “Death to America” was never really laughing. But millions of people may never know the difference.