Fears over AI in midterm political ads miss the real threat | Opinion

As the nation hurtles toward the November midterm election, there is palpable anxiety about artificial intelligence in political advertising, and much of it is warranted.

Last month alone, Republicans aired AI-generated ads depicting Texas Democrat and U.S. Senate candidate James Talarico singing about his support for LGBTQ+ youth and gave Michigan Republican Mike Rogers an AI makeover to appear more muscular in his Senate campaign.

Add these to the growing list of examples this year illustrating how quickly AI political content has moved from a future concern to a present reality.

A fake voice can mislead voters. A synthetic image can travel faster than a correction. A campaign that uses AI to deceive people should be held accountable. None of that is complicated.

Much of the public conversation has focused on how AI is transforming political advertising. That focus makes sense. But political ads are only one part of the information ecosystem voters encounter every day.

Content reaches people through recommendation algorithms, influencers, reposts, clips, memes and countless other forms of distribution that operate largely outside the rules governing paid political advertising.

For years, the political conversation about digital campaigning has focused on paid advertising. What campaigns can say. Who they can target. What must be disclosed. Which ads should be allowed to run, and which should not.

That scrutiny is not going away, nor should it. Political ads should be transparent. Campaigns should not be able to mislead voters with deepfakes or other AI-generated fabrications. Platforms have every reason to take this type of deceptive paid political content more seriously.

But paid political ads are also one of the few parts of the online information ecosystem where rules, disclosures and accountability already exist, at least to some degree. They also represent only a small fraction of the campaign-related content online.

Organic content is another story.

A political ad may be reviewed, labeled, archived, restricted and publicly searchable. A misleading organic post can be boosted by an algorithm, copied by influencers, remixed into a video, stripped of context and pushed into millions of feeds before anyone with authority has even decided whether it violates a policy, which is likely laughably weak if it even exists at all.

That imbalance needs far more attention. The most powerful political content is often the least regulated and accountable.

This is where the AI debate gets too small. Yes, someone can use AI or other means to create a deceptive video. But platforms like Facebook and X can decide, invisibly and at enormous scale, that the fake video is worth showing to millions of people.

Behind those recommendations are business incentives that reward attention, engagement and time spent on the platform.

That’s why they reward content that keeps people watching, clicking, sharing and arguing, because it keeps the money rolling in. In politics, that often means the loudest claim beats the most accurate one. Outrage and conspiracies travel. The dry fact-check of the false claim usually does not.

Campaigns understand this problem because they live inside it. A verified political advertiser trying to reach voters with accurate information can face a maze of restrictions, delays, shifting platform rules and reduced targeting capabilities. Meanwhile, an anonymous account pushing a misleading claim or AI-generated deepfake can ride the wave of organic distribution with far less friction.

That imbalance does not make much sense, and campaigns are increasingly taking notice.

As paid political advertising becomes more constrained, campaigns are investing more resources in building or elevating networks of supporters, influencers, content creators and digital volunteers who can generate and amplify content organically.

Spencer Pratt’s recent Los Angeles mayoral campaign highlighted the trend, amplifying AI-generated content made by supporters casting him as a superhero and his opponents as villains, helping fuel his campaign despite an eventual resounding loss.

Some campaigns are even sponsoring influencer content that reaches voters outside traditional paid placements. Others are clipping longer-form appearances, interviews and speeches into bite-sized videos built for maximum exposure, with accuracy too often treated as an afterthought.

Grassroots participation has always been a part of democratic politics. But when these distribution networks are used to spread falsehoods, deceptive AI-generated content or coordinated manipulation, they can become far more powerful than any traditional political advertisement.

If platforms are serious about protecting elections, they should spend less time treating regulated campaign communication as the center of the threat and more time addressing their own systems that give falsehoods their reach, especially after largely abandoning useful safeguards they built during the COVID-19 era to limit the spread of harmful misinformation.

This is especially important at a moment when major technology companies are moving in the wrong direction. Across the industry, platforms have reduced trust and safety investments, scaled back proactive moderation and leaned into a looser “free expression” posture. At the same time, tech executives and companies have been working to maintain access and influence with President Donald Trump and his administration.

The shift is hardly surprising, and one need not believe in conspiracy theories to be concerned. One of Trump’s first acts after returning to office was to target what his administration described as online censorship, and federal agencies have since been enlisted to scrutinize content moderation practices at major technology platforms.

The issue is what happens when companies with enormous control over political information bend to that pressure, becoming less willing to enforce effective rules, less transparent about what they amplify and more attentive to the political costs of angering the Trump administration.

That is not a healthy or neutral information environment.

A smarter approach would be for platforms to focus enforcement on coordinated networks, repeat offenders and narratives that are demonstrably false, generated using deceptive AI and spreading widely. They should be much clearer about how political content is recommended, ranked and amplified. Users should know more about why they are seeing a political post, especially when it has traveled far beyond the account or community where it began.

Platforms should also stop treating all political actors as if they pose the same risk. There is a difference between a verified campaign running a disclosed ad about early voting and an anonymous network flooding feeds with false information. There is a difference between legitimate political speech and coordinated deception.

Much of the public debate to date focuses on Section 230, the federal law that largely protects Big Tech platforms from legal liability for content posted by their users. The law was written 30 years ago for platforms that primarily hosted content. Today’s platforms also curate, rank, recommend and amplify it, giving them far greater influence over what voters actually see.

A bad post is one thing. A bad post algorithmically pushed into millions of feeds can distort public debate at a scale no individual user ever could. That is where policymakers, campaigns, journalists and voters should be looking.

AI is making deception easier. No serious person should dismiss that. But AI is arriving in an information ecosystem that is already incredibly unhealthy. It is being layered on top of one already defined by opaque algorithms, uneven enforcement and platform rules that often constrain the most transparent political speech while leaving the organic chaos largely intact.

So yes, we should worry about deepfakes in political ads. But we should worry even more about missing the forest for the trees.

The real deepfake may be the idea that synthetic ads are the biggest threat to our democracy, when the far more powerful force is the systems that decide what content voters see, over and over again, in the first place.

Mark Jablonowski is a political technologist and CEO of DSPolitical. He previously built tech infrastructure for Barack Obama’s 2008 campaign and served as CTO for the Presidential Inaugural Committee in 2009.

The views expressed in this article are the writer’s own.