Older voters are significantly worse at identifying synthetic media than younger users. Ken Paxton’s political ecosystem may have found the perfect audience for an age-driven digital literacy gap.
By Omar Afra
Something profoundly weird has happened to the visual language of the American right. Open Facebook for ten minutes and eventually you encounter it: Donald Trump with shoulders like a fucking refrigerator, a suspiciously gorgeous Special Forces soldier clutching an American flag, or super villain Muslim terrorists about to accost a young blonde girl. Everyone is glossy with perfectly coiffed hair and flawless skin. There is an endless cacophony of enormous flags and pensive bald eagles. It is political communication by way of a truck-stop T-shirt rack, except now the truck-stop T-shirt rack can generate unlimited video, and in 2026 this shit has become remarkably normalized on the American right.
AI imagery is obviously not exclusively Republican. Democrats use synthetic media too, and anyone pretending otherwise is trying to win an argument rather than understand what is happening. But the asymmetry this election cycle is increasingly difficult to ignore. Reuters reported in March that Republican campaigns were leading the use of political deepfakes in the 2026 midterms, including fabricated videos aimed at Texas Democrat James Talarico. In Texas specifically, The Texas Tribune found AI spreading throughout Republican campaign advertising, including Ken Paxton’s own synthetic video of John Cornyn dancing with Jasmine Crockett. The animation was crude enough that Cornyn and Crockett occasionally appeared to move like characters in a haunted bowling-alley arcade. Paxton published it anyway.

That ugliness is not necessarily a failure of the technology. It may be part of why the technology works. The mistake is assuming an AI political image has to fool everybody. It doesn’t. It merely has to survive long enough in the environment for which it was made, and increasingly some of the most gloriously obvious AI bullshit on the internet appears to have discovered its natural habitat: Facebook pages serving an older conservative audience. The trick-fuck is not that every viewer is completely bamboozled into believing every synthetic image is documentary evidence. The trick-fuck is that the image only has to be plausible enough, emotionally satisfying enough and fast enough to reinforce whatever the viewer already suspects.

There is an uncomfortable demographic symbiosis here that deserves to be discussed without pretending everyone’s grandmother is an idiot. Older Americans did not grow up learning the visual grammar of synthetic media. Someone who is 25 has spent much of his or her life surrounded by Photoshop, video-game graphics, Snapchat filters, FaceTune, TikTok effects, CGI, memes, face swaps, Midjourney and generative video. That exposure creates something close to an unconscious bullshit detector. The lighting is wrong. The skin has that waxy diffusion. The camera movement makes no physical sense. The crowd in the background looks like twelve people having simultaneous strokes. The body has been chopped up and reassembled according to some approximation of human anatomy. You may not consciously inventory the mistakes. Something simply screams AI.
For somebody whose understanding of photographs was formed in 1975, the relationship between image and reality was historically different. A photograph could lie, obviously, but somebody generally had to put something in front of a camera first. That assumption is now catastrophically obsolete. The result is not an intelligence gap so much as a media-literacy gap, and political operators have discovered that a gap does not need to be enormous to be exploitable. You do not need to completely hoodwink an audience if you can simply get it to stop asking where the picture came from.
Research increasingly shows that this generational difference is real. A large review published in Communications of the ACM found that older people performed worse than younger people at identifying synthetic media across formats. Researchers studying political deepfakes in Nature Communications found accurate detection falling to about 20 percent among participants in their 60s and older in one experiment, compared with higher rates among younger groups. Another study of deepfakes and scientific communication found that every additional year of age was associated with a decline in the likelihood of correctly identifying a deepfake.

Now put that next to Facebook. According to Pew’s latest platform data, 57 percent of Americans 65 and older use Facebook, compared with just 19 percent who use Instagram and 12 percent who use TikTok. Among Americans 50 to 64, Facebook usage is 74 percent. Facebook is therefore not merely another social network in this story. It is an enormous political distribution pipe connected directly to an electorate whose media habits were formed before the invention of the personal computer.
And we already know something disturbing about what travels through that pipe. Researchers examining actual Facebook behavior during the 2016 election found that users older than 65 shared nearly seven times as many links from fake-news domains as the youngest users. The age effect remained even after controlling for ideology, education and partisanship. Conservatives also shared more fake-news material in that particular election, when the fake-news ecosystem was overwhelmingly pro-Trump. That does not mean Republican senior citizens are uniquely stupid. It means somebody has finally built an industrial propaganda machine exquisitely suited to a particular information vulnerability.
The modern right-wing AI ecosystem is practically engineered around it. The content is inexpensive. It is instantly comprehensible. It requires no reading. It thrives on Facebook. It rewards emotional certainty over inspection. It depicts the world not as it exists but as the viewer already believes it exists: Trump is impossibly strong, American soldiers are impossibly heroic, immigrants are impossibly sinister, liberals are impossibly insane and God has apparently hired an extremely aggressive digital art director. Truth barely enters into the transaction because the image is not really asking to be evaluated as evidence. It is asking to be recognized as tribal scripture.
And sometimes even the people producing and benefiting from this ecosystem get fucking bamboozled by it. In April, after an American airman was rescued during the U.S. war with Iran, a synthetic image appeared online depicting the supposed serviceman smiling aboard an aircraft surrounded by troops and clutching an American flag. It looked like an AI generator had been instructed to produce AMERICA FUCK YEAH EASTER MIRACLE 8K CINEMATIC. Uniform details were wrong. A flag patch appeared in the wrong place. One hand appeared to contain an extra finger. Equipment melted into clothing. Detection software later assessed the picture as overwhelmingly likely to be synthetic.
Ken Paxton shared it as real, as did Texas Gov. Greg Abbott and Rep. Mike Lawler. Paxton wrapped the fake image in religious meaning, describing the rescue around Easter as evidence of divine providence, before deleting his post once the fabrication was exposed. (military.com) There, in miniature, is the whole fucking feedback loop: the AI image was created because somebody understood exactly which symbols would make a conservative audience stop scrolling — soldier, flag, rescue, God, Easter, victory — and Paxton encountered those symbols, experienced the intended emotional response and transmitted the image to his own audience without apparently stopping to determine whether the photograph depicted an actual human being. The guy was not simply moving the slop downstream. He got hoodwinked by the same slop himself.
That is what makes the Paxton ecosystem an unusually useful case study. The important question is no longer whether artificial intelligence can create political propaganda convincing enough to survive forensic inspection. Much of the material circulating around conservative Facebook does not come remotely close. Sometimes it looks like absolute dogshit. The more interesting question is why it needs to look any better when the message already has a perfectly matched audience, a frictionless distribution system and an emotional payload engineered to bypass scrutiny.
Political propaganda has historically improved by becoming more persuasive. AI propaganda may have discovered another route: find an audience predisposed to believe the emotional proposition, manufacture unlimited imagery confirming it, place that imagery inside a platform disproportionately populated by older users, and allow repetition and tribal identification to do the rest. At that point the badly rendered hand, the chopped-up anatomy or the waxy face is not necessarily fatal to the operation. The image can be obvious bullshit to one part of the population and still bamboozle, trick-fuck or simply emotionally prime another. In that system, a six-fingered soldier is not necessarily a bug. He may be product-market fit.

Comments