Editor’s note: The Texas Tribune does not use AI to create news content. Our policy prohibits publishing news photos and videos created or modified using generative AI. If the AI-generated image itself is the subject of your story, clearly label it with both a watermark and a caption.
Texas voters heading to the polls in the 2026 primary are encountering a campaign shaped by artificial intelligence, as candidates use it to mock their opponents and dramatize political attacks.
A bill introduced in the Texas state legislature last year that would require disclosure of information about political ads that use AI or that significantly alter real-world images passed the Texas House of Representatives, but stalled in the Senate and was never enacted. This means that AI in political advertising is still free to use, as the technology is rapidly becoming more accessible and practical. Across the country, the use of AI in campaigns has ranged from animated satire to sophisticated deepfakes, testing both the limits of political discourse and voters’ ability to discern fact from digital fiction.
Election disinformation is not new. Cheapfakes, out-of-context, and edited images and videos have become part of disinformation in recent years. AI-generated content is now one of the faster and more accessible tools for spreading disinformation and misinformation.
“In the long run, AI-generated videos and images will desensitize voters to the importance of factually accurate information,” said Ernest Calbo, director of the Interdisciplinary Laboratory for Computational Social Sciences and professor of government and political science at the University of Maryland. “Even for the most knowledgeable actors, it is becoming increasingly difficult to distinguish between the truth and the lies among a vast number of videos and images.” To understand the scope of this trend, The Texas Tribune examined the past five months of Facebook and YouTube ads from candidates running for high-profile offices in the March primaries.
Here’s how candidates are using AI in political ads. Low-tech AI videos mock political opponents and attack party members
In mid-January, Attorney General Ken Paxton, who is running for the U.S. Senate, shared an AI-powered video of his Republican front-runner, Sen. John Cornyn, dancing with Rep. Jasmine Crockett, D-Dallas. “New Ad: John Cornyn dances late into the night with liberal lunatics like Jasmine Crockett and sells us out every step of the way. That’s why he calls Crockett his ‘dance partner’ and she says Cornyn is her ‘best partner’ in the Senate,” he wrote on social media.
Crockett’s campaign told KXAN, “This ad is not substantively attacking her, but attacking her on a bipartisan basis, which most voters appreciate.” “Paxton’s AI-generated digital ad attacks Congressman Crockett, who worked across the aisle to provide solutions to the opioid crisis, stop fentanyl deaths, and address hunger in Texas.”
The video Paxton published on his social media accounts includes a disclaimer about AI-generated content at the end of the video. However, there are also clear signs that it was generated by AI, such as unnatural movements of participants, faceless humans in the background of the video, and blurry backgrounds. It definitely looks like an anime.
Angie Horan, director of the International Fact-Checking Network, said even clearly fake or parody AI poses a risk.
“Those of us who fact-check content online know that many people mistake parodies for real content,” Horan says. “They’re busy, they’re scrolling fast, and they don’t see any clues that what they’re seeing isn’t real. Those who create fake ads are often satisfied with being taken literally.” These types of videos are also making their way into down-ballot elections, including in Texas-level congressional races. Cat Wall, a Republican candidate running for a state House seat, released a fake YouTube video satirizing her opponent, Texas Rep. Angelia Orr (R-Itasca), that uses AI to display deepfakes of Vladimir Putin and Xi Jinping. The video features a synthetic voice clone and manipulated images of the candidate, mocking her “voting record” and making her appear to be an ally of a foreign adversary.
At the end, the narrator says, “This ad is a parody using AI video tools.” There are several indicators in the content that indicate that the video was generated using AI. For example, there is no synchronization between the lips of the person representing Orr and her character’s voice.
Additionally, the outfit and setting are very similar to an interview she gave to the Waco Tribune-Herald four years ago. However, the footage shows notable differences, such as the shape of the teeth, which are characteristic of AI-generated videos. A Wall spokesperson acknowledged that AI could be a weapon in deceptive campaigning, but said their campaign had “safeguards” in place, including clarifying the use of AI in clear language, ensuring that AI is only used to present points that can be proven in writing, and citing sources for viewers to “find documentary evidence.” Some of the AI videos reviewed show obvious cartoons. On December 11, 2025, U.S. Representative Jasmine Crockett posted an AI video of herself as a baby on her Facebook page, accusing Baby Trump of cheating in the Texas election. This content was published by Diaper Diplomacy, a political parody YouTube channel that uses AI to portray public figures as childlike. The platform said that on YouTube, videos are labeled as “altered or synthetic content,” meaning that “the audio or video has been heavily edited or digitally generated.”
Sophisticated AI satirical video
On January 21, Mr. Cornyn posted a fake AI-generated video depicting Congressman Wesley Hunt as a “show dog.” Unlike other AI ads, the attack on Wesley Hunt showed a much more sophisticated level of digital manipulation and did not disclose any use of the technology. However, some specialized AI detection tools, such as Hive Moderation and Google’s AI chatbot Gemini, showed that there was a 99% chance that the video was created using AI.
While the tool is not 100% reliable and may introduce errors, there are certain visual elements that suggest the video was generated using AI. The blurry sign in the background of the opening frame, the American flag folded in an unusual way, the somewhat unnatural movements of the person meant to represent Hunt, and the overly polished quality of the video.
“Technology moves so fast,” says PolitiFact editor-in-chief Katie Saunders. “What could happen is that people become even more skeptical of what they’re seeing. If they don’t feel like they can believe what they’re seeing, they might stop believing in anything.”
On February 8, Crockett released an ad titled “Texans don’t back down. We rise up.” In the ad, Crockett describes himself as a “warrior” and “champion” for Texans’ voting and rights, while also characterizing the current system as “broken” and communities “in fear.”
Gemini said the final image of the video, which shows Crockett surrounded by dozens of people, was created using Google’s AI.
Also, as sometimes happens with this kind of content, and I can’t confirm with 100% certainty due to the quality of the video, some faces and some body outlines in the image are blurred and not clearly visible, unlike Crockett’s figure and the people around her.
Crockett’s ad was criticized on social media for portraying supporters as AI. Crockett’s campaign responded, but did not say whether AI was used.
“Our new ad is the result of hundreds of hours of hands-on production and collaboration between creators and unions. This ad is a deliberate and meaningful attempt to connect with and energize new voters,” her campaign responded. Crockett, Cornyn and Paxton’s campaigns either did not respond or declined to comment to the Tribune.
Cornyn has posted several AI-generated images depicting Democrats, including state Rep. James Talarico of Austin and former U.S. Rep. Beto O’Rourke. In a series of AI-generated images and videos, Talarico, who is running for the U.S. Senate, transforms into Taxra, a caricature of a fanged vampire in a suit, clutching a “tax bill” with distorted, AI-distorted fingers.
Cornyn also released an AI-generated image of O’Rourke depicting “FrankenBeto” as a monster driven by the “California Mandate” and “chaos” attacking his energy policies. And there was an AI-generated “Halloween Special” graphic depicting former U.S. Rep. Colin Allred of Dallas as a green-skinned witch boiling a cauldron of “Bidenomics” in front of the dark, storm-ravaged Texas Capitol.
Experts say that if you see suspicious content on social media, the golden rule is to think twice before sharing it.
“Ask yourself: Where did this come from? Is this from an official campaign source? Did some news organization verify it?” Horan said. “If you think something is intended to stir up emotion or anger, treat it as a warning sign and watch carefully before sharing. Also, be more intentional about where you get your information. Rely on established news sources, local journalists who know the candidates well, and verified campaign communications rather than content that appears in your feed.”
Disclosure: Facebook and Google have financially supported The Texas Tribune, a nonprofit, nonpartisan news organization that is funded in part by contributions from its members, foundations, and corporate sponsors. Financial supporters play no role in the Tribune’s journalism. See the complete list of them here.
Source link
