How to Tell If a Political Campaign Ad Video Is AI Generated in 2026

How to tell if a political campaign ad video is AI generated in 2026? Look for the label first, the original second, and the payer third — and do not trust your eyes alone. The scale of the problem is now measurable. The Wesleyan Media Project, which has tracked political advertising since 2010, counted at least 220 ads in the 2026 US midterms that were made or enhanced with AI, backed by more than $137 million in spending — and only 27 percent of them carried any AI disclaimer. The midterms were held on November 3, 2026, but these checks still matter: special elections, primaries, and the next cycle are already coming.

The hardest part is that an AI campaign ad is often legal: using AI in a political ad is legal in itself. What 30 US states regulate is fake media that makes a real person — usually a candidate — appear to say or do something they never did. Most want a label, not a ban. So the question is rarely “is this allowed?” It is “is this real, and who made it?”

Check 1: hunt for the label

Where a state requires a disclosure, it is usually short on-screen text, or a line read aloud at the start and end of an audio ad. Look for wording like this:

• “Created in whole or in part with the use of generative artificial intelligence (AI).” — Florida
• “This video content generated by AI.” — Wisconsin and Utah
• “Ad generated or substantially altered using artificial intelligence.” — California
• “This image has been manipulated.” — Nevada, New York, and Washington

Some states require the label on screen for the whole video; Florida, Maine, and Montana set a minimum size of 4 percent of the picture height — small on a phone. Pause the ad and scan the corners and the bottom of the frame, especially in the first few seconds. Labels are easy to miss: the National Republican Senatorial Committee’s March 2026 ad featuring an AI-made version of Texas Senate candidate James Talarico did carry a label, but it was faint text in the bottom right corner of an 85-second video. A Berkeley expert on fake media, Hany Farid, said the face and voice in that ad were “very good” and doubted most people would notice it was fake.

A label tells you the picture or voice was made or changed with AI — not that the claim is false. And no label tells you nothing at all: only about one in four AI ads tracked this cycle carried any disclaimer.

Check 2: find the original

If a candidate seems to say or do something wild, check the candidate’s own verified channels first, then coverage from a news outlet you know. A real moment that big will be everywhere; a fake usually lives only in the ad.

Then check who paid for it. Every ad on Facebook and Instagram is searchable in the Meta Ad Library, and Google and YouTube ads are in the Google Ads Transparency Center. Both show the sponsor, and Meta shows the spend. The “paid for by” line on the ad itself tells you who is responsible for it.

A reverse image search helps too. Take a screenshot of one frame and run it through Google Lens or TinEye to see where else the image appeared and when. Real clips usually trace back to a speech, an interview, or a debate. If the only source is the ad itself, that is a red flag.

Check 3: watch the face and listen to the voice

Visual tells are weaker than they used to be — the best 2026 tools fixed most of them — but they still work as supporting evidence. In the Talarico ad, the one flaw the Berkeley expert noticed was lip-sync: lips that did not quite match the voice. Watch for that, plus odd blinking, mismatched reflections, teeth that blur or change shape, and plastic-smooth skin.

Listen as carefully as you look. Cloned voices can sound robotic, pause in unnatural places, or carry no breaths and no room sound. A candidate’s voice that is perfectly even over footage with no matching ambient noise should raise suspicion.

And check your pulse. If a clip makes you furious in five seconds flat, that is the point. Outrage spreads, and fakes are built to get a quick reaction. Slow down before you hit share.

What the law says — and where to report

Thirty states now have an enforceable law aimed at AI-generated or digitally faked content in election ads. Before 2024 only five did. Twenty-six require a disclosure; three — Maryland, Minnesota, and Texas — ban certain election deepfakes outright; California does both.

Enforcement is uneven. In about a quarter of them, the only remedy is a lawsuit by the candidate who was faked. Where an agency does enforce, it is usually the Secretary of State, an ethics or campaign finance commission, or a prosecutor. One case from this cycle: on October 23, 2026, the Mississippi Attorney General wrote that an unlabeled AI ad from Democrat Scott Colom’s campaign, showing Senator Cindy Hyde-Smith, appeared to violate the state’s deepfake law — and the campaign deleted it the same day.

Before you report, save the evidence, because online ads disappear fast: a screenshot or screen recording of the ad, the link to where you saw it, the date and time, the platform or station, the “paid for by” line word for word, and the ad library entry if there is one.

Frequently asked questions

Is it illegal for a campaign to use AI in an ad?

No. What the law targets is deceptive media: a fake video of a real person saying or doing something they never did, usually without a required label. Most states allow the ad if it carries the disclosure; Maryland, Minnesota, and Texas ban certain election deepfakes outright.

Can an AI video detector prove a campaign ad is fake?

No detector can prove it. A tool like Google’s SynthID Detector or https://aivideodetector.site/ can add evidence about a particular clip, but a clean result never proves an ad is real, and a flagged result does not prove the claim in the ad is false.

What if the ad is satire or a parody?

Satire is where most legal challenges land. Courts have pushed back on the broadest bans — Montana’s law was partly blocked in September 2026, Hawaii’s was struck down in January 2026, and California’s broadest ban fell in 2025. But satire does not excuse deception: an ad that presents a fake clip as real, with no label and no parody framing, is exactly what the disclosure laws target.

How is this different from your guide to campaign AI rules?

Our earlier piece explained whether campaigns are allowed to use AI videos in ads. This guide is the voter-side companion: how to spot one, verify it, and report it if it breaks the law where you live.