If you feel like it is getting harder to tell what is real in politics, you are not imagining it. In just a few years, AI-generated political ads have gone from fringe experiments to a normal part of campaign arsenals around the world. Campaigns can now spin up photo-realistic videos, fake audio of opponents, and personalized message variations in minutes, at a fraction of the old cost.

That speed and scale change everything. You are no longer just watching an ad that a consultant storyboarded weeks ago; you may be seeing content that was generated this morning, A/B-tested on social media by lunchtime, and micro-targeted at your demographic by dinner. As one academic study of AI and elections put it, the falling cost of creating synthetic media makes it far easier to run disinformation operations at scale using deepfakes and other AI-generated content (Examining the Implications of Deepfakes for Election Integrity).

At the same time, regulators and platforms are scrambling to respond. The US Federal Election Commission (FEC), the Federal Communications Commission (FCC), and the European Union have all moved – in very different ways – to clarify how existing laws apply when the ad you see is built by an algorithm. You are watching a new campaign frontier emerge in real time.

What Do We Mean by AI-Generated Political Ads?

Let’s get concrete about what counts as an “AI-generated” ad, because it is not just one thing.

Today’s political ads can use AI at several layers:

  • Synthetic visuals: Image and video generators like Midjourney, DALL·E, Runway, and Pika Labs can create crowds, protests, disasters, or glossy candidate imagery that never actually happened.
  • Synthetic voices: Voice-cloning tools can make a candidate (or their opponent) say anything in a realistic voice, in multiple languages and tones.
  • Script and copy generation: Large language models (LLMs) like ChatGPT, Claude, and Google Gemini can draft ad scripts, fundraising pitches, or micro-targeted variations for different audiences.
  • Targeting and optimization: AI systems can test hundreds of versions of an ad and optimize which one you see based on your behavior, location, and engagement history.

Some of this is relatively benign and even helpful: for example, using AI to automatically translate an ad into multiple languages or cheaply produce explainer animations. But the same tools can be used to fabricate “news clips” of imaginary events, or fake audio of a candidate admitting to crimes.

Real-World Examples: From Dystopian Futures to Deepfake Voices

You have already seen some of this, even if you did not realize it.

In April 2023, the US Republican National Committee released a video ad attacking President Joe Biden that was made entirely from AI-generated images. The ad imagined a dystopian second Biden term with fabricated scenes of international crisis and domestic chaos (deeplearning.ai summary of the RNC ad). This was widely cited as one of the first major party US campaign ads built wholly from generative imagery.

Not long after, a pro-Ron DeSantis group released an ad that used AI to clone Donald Trump’s voice reading posts he had written about Dr. Anthony Fauci. PolitiFact analyzed the spot, noting that it included three AI-generated photos of Trump with Fauci and that the voice track was described by the sponsor as “AI-generated” rather than authentic audio (PolitiFact on the AI Trump voice ad). That is a decisive shift: political ads are no longer limited to real footage and real sound.

Outside the US, campaigns have woven AI into their visual branding too. In Indonesia’s 2024 presidential race, candidate Prabowo Subianto’s campaign used AI-generated art extensively, including cute cartoon-like portrayals of him as a child to soften his strongman image (overview of AI and elections).

These cases are important for you not because they are the only examples, but because they show how quickly the norms are shifting. What was shocking in 2023 – a fully synthetic attack ad – is likely to be normal in the next election cycle.

Why Campaigns Love AI: Speed, Scale, and Micro-Targeting

From a strategist’s point of view, AI-generated ads are irresistible. They offer:

  • Lower production costs: Instead of hiring actors, renting locations, and editing footage for weeks, a team can prompt a video generator and refine the output in hours.
  • Instant message testing: LLMs can spin dozens of script variations (“make this angrier,” “make this appeal to small business owners”) that can be tested and optimized on social platforms.
  • Hyper-personalization: AI can swap out imagery, phrasing, even accents, so that different demographic segments see different versions tailored to their fears or hopes.
  • 24/7 content machine: Campaigns can keep feeds populated with memes, short clips, and “news-style” videos without burning out human creatives.

A recent research paper on “AI-enabled campaigning” suggests grouping these uses into three broad buckets: campaign operations, voter outreach, and deception. The same underlying tools can streamline back-office tasks, supercharge targeting, or be used to mislead and manipulate voters (Deceptive uses of Artificial Intelligence in elections).

When you hear about campaigns using ChatGPT, Claude, or Gemini, most of the time it is in those first two buckets: writing emails, segmenting lists, translating speeches. It is the third bucket – deception – that keeps election officials up at night.

The Dark Side: Deepfakes and “AI Slop”

The nightmare scenario is the late-breaking deepfake: a realistic fake video or audio clip of a candidate saying something racist, corrupt, or incoherent, released just days or hours before voting. There may be no time to debunk it before it goes viral.

Researchers warn that deepfakes are already being used in elections to both reinforce existing narratives and create adversarial attacks that distort public perception (deepfake and election integrity study). Generative models are getting better every year, and the cost barrier is falling.

On the lower end of the quality spectrum, there is also what some commentators call “AI slop” – low-effort, low-quality AI images and videos pumped out purely to game social media algorithms and flood the zone with emotionally charged but misleading content. Analyses of the 2024 US election cycle have noted that cheap synthetic memes and crudely edited clips are now routine in certain partisan ecosystems (AI-generated content in American politics).

This creates a few big problems for you as a voter:

  1. You may see false or heavily manipulated content that looks real at a glance.
  2. Even real footage can be dismissed as “probably AI,” eroding trust in authentic evidence.
  3. The sheer volume of content makes it harder for fact-checkers and journalists to keep up.

How Regulators Are Responding in the US

In the United States, the legal response is still very much in progress – and fragmented.

  • The Federal Election Commission (FEC) enforces campaign finance laws, including rules against fraudulent misrepresentation in campaign communications. In 2024, the FEC adopted an interpretive rule clarifying that an existing federal law against misrepresenting yourself as speaking or acting on behalf of a candidate or party applies “irrespective of the technology being used” – which includes AI tools (FEC interpretive rule on AI in campaign ads). In plain language: using deepfakes to impersonate a candidate is still illegal, even if it is done with generative AI.

  • The Federal Communications Commission (FCC), which regulates broadcast media, has gone a step further. In July 2024, it announced a proposal to require that political ads on TV and radio disclose when they contain AI-generated content, so that viewers and listeners are “fully informed” when synthetic media is used (FCC news release on AI transparency in political ads). The details are still being debated, but the direction is clear: expect on-air labels for AI content.

  • Meanwhile, individual US states have started passing their own laws restricting deceptive deepfake ads — often focused on bans or civil penalties for fake audio or video of candidates within a certain window before an election. The result is a patchwork where the rules differ depending on where you live.

None of this, so far, directly regulates how campaigns can use AI simply to optimize, personalize, or cheaply produce ads. The line regulators are trying to draw is between using AI as a production tool and using AI to deceive – and that line is not always obvious in practice.

The EU’s Transparency-First Approach

If you look beyond the US, the European Union has taken a broader, more harmonized path.

In March 2024, the EU formally adopted Regulation (EU) 2024/900 on the transparency and targeting of political advertising. Among other things, it requires that political ads be accompanied by a transparency notice indicating who paid for them and, crucially for AI, whether “the advertisement contains political advertising that uses techniques involving personal data to target specific individuals or groups” and information on the use of targeting and amplification tools (EU press release on the new regulation; see also the full legal text summarized at EUR-Lex).

While the law does not single out AI by name in every provision, it squarely targets the profiling and micro-targeting capabilities that AI supercharges. It pushes for:

  • Clear labeling of political ads.
  • Public information about who is being targeted and why.
  • Stronger oversight of ad libraries and data use.

For you as a European voter, this should gradually mean more visible labels, more consistent transparency across platforms, and better tools to see who is trying to influence you and how.

Where Platforms and AI Makers Fit In

There is also a third layer beyond governments and campaigns: platforms and AI providers themselves.

  • Major social networks have begun requiring labels for “synthetic or manipulated media” in political contexts and in some cases banning realistic deepfakes of private individuals.
  • AI model providers, including OpenAI (ChatGPT), Anthropic (Claude), and Google (Gemini), have introduced specific policies restricting the generation of targeted political persuasion, especially about contemporary elections. They often block prompts that explicitly ask for tailored messaging at a specific voter group, even though more generic political content is still allowed.

These measures will not catch everything – you can always download an open-source model and run it locally – but they set some norms and make the laziest forms of abuse harder.

For you as a user, it means two things: you are less likely to see obviously manipulative political ad scripts coming straight out of mainstream chatbots, but you will still need to assume that many of the memes and videos in your feed have passed through AI systems somewhere in the pipeline.

How You Can Protect Yourself as a Voter

You cannot single-handedly fix campaign finance law. But you do not have to be helpless, either.

Here are some practical steps you can take:

  1. Adopt a “trust, then verify” mindset. If you see a shocking clip of a candidate saying something outrageous, especially close to an election, pause. Check whether reputable outlets or fact-checkers (AP, Reuters, major national outlets, specialized sites like PolitiFact) have reported on it.

  2. Look for context and source. Who posted the ad? Is it from an official campaign account, a PAC, a random anonymous page, or a meme account? Platforms increasingly label political advertisers; use those labels.

  3. Use reverse searches and basic forensics. Tools like Google reverse image search or InVID for video can sometimes show whether an image or clip has been altered or is brand-new.

  4. Pay attention to disclosures. As regulations like the FCC proposal and the EU’s 2024/900 regime roll out, you will start to see more explicit labels when AI is used. They are imperfect, but they are signals you should notice, not ignore.

  5. Diversify your information diet. If all of your political content comes from highly partisan feeds, AI will happily tailor ever more extreme content to your tastes. Following a mix of perspectives and some non-partisan sources makes you harder to micro-target and manipulate.

Conclusion: This Frontier Is Here – Now Learn to Navigate It

AI-generated political ads are not a future problem; they are already shaping elections today. Campaigns love them because they are cheap, flexible, and scalable. Regulators are trying – unevenly – to demand transparency and curb outright deception. And you are caught in the middle, scrolling past videos that may or may not show anything that ever happened.

To stay grounded in this new campaign frontier, you can:

  1. Pick one or two trusted, non-partisan news or fact-checking outlets and make a habit of checking them whenever an ad or clip really riles you up.
  2. Spend five minutes today learning one verification trick – like reverse image search – and use it at least once during the next big political news cycle.
  3. Talk about AI-generated ads with friends and family, especially those who are less online; simply knowing that deepfakes and synthetic ads exist makes people less likely to be fooled by the first one they see.

You do not need to become an AI expert to protect yourself. You just need to assume that, in modern politics, some fraction of what you see is synthetic – and act accordingly.