I have written about grift before. The grift is the con at scale, the confidence game run from the highest office in the land, open and defended as virtue. This essay is about the machinery that makes the grift possible: a presidency that no longer argues with reality but generates its own.

PolitiFact counted 36 posts on Donald Trump’s Truth Social account since his inauguration that used AI-generated imagery. Twenty-one of them promoted or sensationalized Trump and his administration. Twelve degraded his opponents. Three reinforced his policy messaging (PolitiFact, 2025). NBC News, reviewing the same feed, found “dozens” of pieces of synthetic media, roughly half of them posted in August and September 2025 alone (NBC News, 2025). The White House’s own X account has posted at least 14 more since January. Add the Department of Homeland Security, the Social Security Administration, and other agency accounts, and the count grows further.

This is a presidential first. No prior administration used generative AI as an everyday instrument of governance. Katherine Ognyanova, a communications professor at Rutgers, put it plainly: “While presidents have been known to distort reality in the past, there is no precedent for this regular dissemination of deepfakes from the Oval Office” (PolitiFact, 2025).

The anatomy of a synthetic presidency

The posts fall into three categories, and each category serves a different function.

First, self-glorification. The bulk of the AI content transforms Trump into whatever the moment requires: the pope, less than two weeks after Pope Francis died; Superman; a Jedi; a king on a Time magazine-style cover reading “Long Live The King”; a Vietnam War commander in an Apocalypse Now spoof captioned “I love the smell of deportations in the morning” (PolitiFact, 2025). He roars alongside a lion. He conducts an orchestra at the Kennedy Center. He stands at a stock exchange bell beside a sign for a company that does not exist. On July 4, his account shared a video of his own face transforming from baby to adult, and when asked about it three days later, he said: “It was AI-generated. So, it’s a little bit scary” (PolitiFact, 2025).

Second, attacks on opponents. This is where the synthetic media does its sharpest work. On September 29, 2025, Trump posted a deepfake video set to mariachi music depicting House Minority Leader Hakeem Jeffries in a sombrero with a fake mustache and Senator Chuck Schumer saying “Nobody likes Democrats.” It was reshared 19,400 times and provided “satirical packaging” for the false Republican talking point that Democrats shut down the government to give immigrants free health care (PolitiFact, 2025). A September 8 deepfake used a falsified voice to depict Nancy Pelosi admitting to insider trading. A July post showed Senator Adam Schiff in handcuffs. A September 3 video depicted Governor JB Pritzker and former Governor Chris Christie as sumo wrestlers (PolitiFact, 2025; NBC News, 2025).

Third, policy signaling. Some posts are not jokes at all. When critics questioned the $400 million jet from Qatar, Trump’s account shared an AI image of the Statue of Liberty wearing a sign that read “gift from a foreign nation” (PolitiFact, 2025). In August, he posted a fictitious image of himself ringing the New York Stock Exchange bell beside a sign for the “Great American Mortgage Corporation,” ticker MAGA, a company that does not appear to exist but which referenced real discussions about selling the government’s stakes in Fannie Mae and Freddie Mac (NBC News, 2025). A cartoon showed him firing Federal Reserve Chairman Jerome Powell (NBC News, 2025). These posts are policy announcements without the inconvenience of policy: a picture that says what a statement cannot yet say, retractable when the market or the court objects.

The limit case

On July 20, 2025, Trump posted a video to Truth Social showing former President Barack Obama being arrested in the Oval Office. The video opened with real clips of Democratic officials saying “no one is above the law,” then cut to an AI-generated scene of FBI agents in sunglasses grabbing Obama and dragging him to the floor, followed by fabricated footage of Obama in an orange prison jumpsuit behind bars, all set to “Y.M.C.A.” (Snopes, 2025; The Hill, 2025).

The timing was not accidental. The post appeared two days after Director of National Intelligence Tulsi Gabbard announced she was referring Obama administration officials to the Justice Department for possible criminal prosecution, and near the height of the MAGA backlash over the administration’s handling of the Jeffrey Epstein files (Snopes, 2025). The video was a fantasy of retribution posted at the moment the fantasy was being entertained by the government itself.

The White House defended it. Deputy Press Secretary Harrison Fields said: “President Trump was right about the Obama-Biden administration’s clear involvement in the greatest witch hunt in American history… The President and his entire administration are committed to unearthing wrongdoing and holding any individual accountable for this gross abuse of power and blatant conspiracy against President Trump and his supporters” (Snopes, 2025).

Read that statement again. The defense of a fabricated video of a former president’s arrest is not “it was a joke” or “it was satire.” It is a statement of intent: the administration is committed to holding people accountable, and the video is evidence of what accountability looks like. The deepfake is not a meme. It is a threat with a laugh track.

This is the limit case because it crosses a line no president has crossed: depicting a predecessor in a fabricated criminal scenario, posted from the presidential account, defended by the presidential press office. The Obama spokesperson called Trump’s claims about the 2016 and 2020 elections “bizarre” and “ridiculous” (NBC News, 2025). But the video had already been seen. The denial arrives after the image.

The accusation machine

In August 2024, during the campaign, Trump posted on Truth Social that Kamala Harris had “CHEATED at the airport” and that a photograph of her Detroit rally crowd was “A.I.’d” (PolitiFact, 2024). “There was nobody at the plane,” he wrote. The attendees “didn’t exist.”

The claim was rated Pants on Fire. The photograph was taken by a Harris campaign staffer on an iPhone, corroborated by dozens of photos and videos from Getty Images, the Associated Press, and local news outlets, and analyzed by a UC Berkeley deepfake-detection expert who found no evidence of AI generation or manipulation (PolitiFact, 2024). Local media estimated 15,000 people attended (BBC, 2024; CBS News, 2024).

This is the other half of the machine. Trump does not merely deploy synthetic media to glorify himself and attack opponents. He deploys skepticism about synthetic media as a weapon against opponents, training his audience to disbelieve any image that does not flatter him while believing any image that does. The Harris crowd was real, and he called it fake. The Obama arrest was fake, and he posted it as real. The asymmetry is the strategy: AI is whatever serves the narrative of the moment.

Samuel Woolley, a communications professor at the University of Pittsburgh, said that generative AI images and video “bend reality” for the people who consume them, serving “the goals of individuals and entities that are hoping to create a narrative that is alternative to scientific facts” (NBC News, 2025). The bending is the point. A public that cannot trust any image cannot trust any evidence, and a public that cannot trust evidence cannot hold power accountable.

The defenses, and why they work

Every time a Trump AI post draws criticism, the response follows a script.

Press Secretary Karoline Leavitt, asked about the “medbeds” conspiracy video: “I think the president saw the video and posted it and then took it down. And he has the right to do that. It’s his social media” (NBC News, 2025). The platform is his. The accountability is optional.

Leavitt again, on AI posts generally: “He likes to share memes. He likes to share videos. He likes to repost things that he sees other people post on social media as well” (PolitiFact, 2025). The president is just a guy sharing memes, like anyone else with a phone.

The White House X account, July 11: “Nowhere in the Constitution does it say we can’t post banger memes” (PolitiFact, 2025). July 20: “The memes will continue” (PolitiFact, 2025).

House Speaker Mike Johnson, asked about the video of Trump dumping feces on “No Kings” protesters: “The president uses social media to make a point. You can argue he’s probably the most effective person who’s ever used social media for that. He is using satire to make a point” (PolitiFact, 2025).

Trump himself, on the AI image of himself as pope: “They can’t take a joke?… Give me a break. Somebody did it in fun” (PolitiFact, 2025; NBC News, 2025).

Each defense is a different exit from the same room. It was a joke. It was satire. It was somebody else. It was just a meme. The president is transparent. The defenses are designed to be mutually exclusive and individually sufficient: if one fails, another is available. And because the posts are fun, because they work, because they generate the attention that is the currency of this presidency, the defenses are never tested. They are performed.

Valerie Wirtschafter, a fellow at the Brookings Institution who studies AI in politics, described the function plainly: “They go viral. They get attention. They’re this meme-ification of politics… They’re a little odd. But they get shared. They get eyeballs” (NBC News, 2025).

The policy expression

The synthetic presidency is not separate from the AI policy. It is the same instinct expressed in two registers.

On July 23, 2025, the same week the AI Action Plan was released, Trump signed an executive order titled “Preventing Woke AI in the Federal Government.” It requires that large language models procured by the federal government be “truth-seeking” and “ideologically neutral,” and it defines the failure of neutrality to include “the suppression or distortion of factual information about race or sex” and “concepts like critical race theory, transgenderism, unconscious bias, intersectionality, and systemic racism” (The White House, 2025). The order’s examples are telling: an AI model that “changed the race or sex of historical figures” when asked for images, and one that “asserted that a user should not ‘misgender’ another person even if necessary to stop a nuclear apocalypse” (The White House, 2025).

The order does not regulate AI. It defines whose reality AI is allowed to reflect. “Truth-seeking” in this administration means truth that serves the administration. “Ideological neutrality” means neutrality toward the administration’s ideology. The same president who posts deepfakes of his opponents in handcuffs signs an order demanding that government AI be free from ideological bias. The bias he objects to is not the bias in his own feed. It is the bias in any tool that might disagree with him.

And the president’s own AI product illustrates the limit. In August 2025, Trump Media launched Truth Search AI on Truth Social, powered by Perplexity. When users asked it factual questions, it contradicted the president: tariffs are taxes on American consumers; the 2020 election was not stolen; January 6 was an insurrection linked to Trump’s false claims; grocery prices have not come down; Trump has not ended any wars (Snopes, 2025a). The product’s sources were restricted almost entirely to conservative outlets: Fox News, Fox Business, The Washington Times, The Epoch Times, Breitbart, Newsmax (TechCrunch, 2025; Axios, 2025). Even the president’s own AI, drawing from the president’s own approved sources, produced answers the president did not like. The fix was not to expand the sources. It was to restrict them further.

The audience is not stupid

Here is what the critics of this presidency keep getting wrong, and what the deepfake presidency understands perfectly: the audience is not being deceived.

The people who share the Jeffries sombrero video do not believe Hakeem Jeffries wore a sombrero. The people who watched the Obama arrest video do not believe Barack Obama was arrested. They know it is fake. What they are buying is not information. It is alignment: the feeling of being inside a reality where the people they dislike are criminals, where the president is strong, where the joke is on the people who take things seriously. The deepfake is not a lie told to a mark. It is a signal sent to a tribe.

Michelle Amazeen, a mass communication professor at Boston University who researches misinformation, said the effect is to dominate the information environment and create “the illusion that this is the way things are now” (NBC News, 2025). The illusion is collaborative. The audience helps build it. Every share is a vote for the reality the deepfake depicts.

This is why fact-checking does not work on this content. A PolitiFact rating of Pants on Fire does not debunk the Obama arrest video for the people who watched it. It confirms it: the elites are trying to hide the truth, the deepfake showed what they will not report, the rating is just another cover-up. The fact-check becomes fuel. The debunk becomes evidence of the conspiracy. The synthetic presidency has learned to weaponize the very institutions designed to correct it.

Amazeen’s research found that 84 percent of respondents in a 2025 Ipsos survey said AI-generated content should be clearly labeled, and 81 percent said social media platforms should be required to remove unauthorized deepfakes. Trump’s AI posts are rarely labeled as such (NBC News, 2025). The public wants guardrails. The president is building a machine that requires their absence.

What the deepfake presidency costs

The costs are not abstract. They are the same costs that The Corruption Tax measured in dollars and Corruption at the Summit measured in constitutional terms: the erosion of shared reality, on which every democratic function depends.

A republic requires citizens to be able to agree on what happened. Election results. Crowd sizes. Whether a video shows an arrest or a fabrication. When the president’s own account is a source of synthetic media at industrial volume, the baseline of shared fact becomes contested by default. Every real photograph becomes suspect, because the president has posted fake ones. Every real scandal becomes deniable, because the president has posted fake ones. The epistemic ground erodes in both directions at once: too much doubt about what is real, and too much certainty about what is not.

The office itself is degraded. The presidency is the one institution whose word is supposed to carry weight precisely because it is used sparingly and carefully. A president who posts AI-generated fantasies of his opponents’ arrests, who shares conspiracy theories about medical beds, who responds to the largest protest in American history with a video of himself wearing a crown, is spending the office’s credibility on content. The content is cheap. The credibility is not. When the next real crisis arrives, the president will need the public to believe him. He will have trained them to doubt.

And the precedent is contagious. Gavin Newsom’s press office has already posted AI-generated videos attacking JD Vance, adopting the same tactics in response (NBC News, 2025). Once the norm is broken, it is broken for everyone. The deepfake presidency does not stay a Trump presidency. It becomes the way presidencies communicate, and the public learns to trust nothing.

The honest word

I keep returning to the image of the marble mask split in two: one half crowned in gold, the other half cracked in darkness. That is the deepfake presidency. The same account posts the fabricated arrest of a former president and a video of the president’s own face aging from infancy to manhood, marveling at the technology that made it. The same office signs an order demanding “truth-seeking” AI while its own AI product contradicts it and its own feed contradicts itself hourly. The same press secretary defends a fabricated criminal scenario with a statement about “holding any individual accountable” and defends a conspiracy video with “it’s his social media.”

The defenses are not contradictions. They are the strategy. A joke is a threat. A meme is a policy announcement. A deepfake is a trial run for an arrest. Every category collapse is deliberate: if everything is a joke, nothing is accountable; if nothing is accountable, everything is permitted. The synthetic presidency has found the one loophole that survives every check: make the fabrication indistinguishable from satire, and the satire indistinguishable from governance.

Daniel Schiff, an assistant professor of technology policy at Purdue who runs a database of political deepfakes, described the effect on the audience: “The plurality of deepfakes are about Trump,” he said, excluding sexual deepfakes. The president is not just the most prolific user of synthetic political media. He is its most popular subject, among fans and critics alike (NBC News, 2025). The deepfake presidency has made the president into the deepfake: a figure so saturated in fabrication that the real man and the synthetic man are the same product.

That is the endgame, and it is worth stating plainly. Not the censorship of AI, which this administration opposes in every form that would constrain it. Not the regulation of deepfakes, which it has not proposed. The endgame is a public sphere in which the truth value of any image is decided by tribal loyalty: my tribe’s deepfakes are satire, your tribe’s deepfakes are fraud, and the referee is the man who posts the most of them.

A republic cannot function in that sphere. It can only survive it by refusing to participate: by declining to share, by demanding labels, by treating the presidential account as what it has become, a content channel with nuclear codes. The office will not reform itself. The feed will not label itself. The only remaining check is a public that remembers what it means to see something and believe it, and that treats the erosion of that capacity as the emergency it is.

The con works only as long as we keep buying. The deepfake works only as long as we keep watching.

Sources

PRH | huffmanwrites.org | © Philip Huffman