Hundreds of fake pro-Trump AI-generated avatars are flooding social media
Social media has become a silent battlefield ahead of the U.S. midterm elections, and the army marching through it isn’t made up of real people.
More than 300 accounts have been identified on TikTok, Instagram, Facebook, and YouTube, all featuring avatars generated by artificial intelligence with one clear goal: look human, gain followers, and push pro-Trump messaging to as many people as possible.
The New York Times has been tracking this phenomenon since January 2026, and what they found is, to put it mildly, disturbing. Some of these accounts have racked up more than 35,000 followers, and certain posts have surpassed half a million views, without a single one being labeled as AI-generated content. And here’s the detail that might grab your attention the most: President Trump himself reposted content from one of these avatars on Truth Social — a platinum blonde figure spreading unsubstantiated claims about the governor of California. 😬
The question that lingers is simple yet terrifying: if even the president shared fabricated content, what can we expect from the average voter scrolling through their feed on a random afternoon?
The videos that look real but are 100% fabricated
The pattern across these accounts is so repetitive that, from the outside looking in, it’s almost obvious. In one TikTok video, a blonde films herself with a group of women on a running track and says: if you support Trump, you just made a new friend. In another, a brunette appears with a group at a stadium and repeats the exact same line. In a third, a redhead is on a basketball court and, you guessed it, says the exact same thing.
Each video also comes with an identical and grammatically awkward caption, something along the lines of: I’m new here and I love God, America, and Trump!! All of these videos are the work of artificial intelligence, and this is just one of the many formats these accounts use to try to appear authentic and drive engagement.
The avatars typically present themselves as ordinary men and women, almost always attractive-looking, staring engagingly into the camera while talking about hot-button topics like the war in Iran, abortion, or even artists like Bad Bunny. The posts are packed with references to the so-called America First movement and what they describe as the radical left. All of it wrapped in a format that perfectly mimics the behavior of real influencers.
The scale of the operation and who’s tracking it
The New York Times began monitoring these pro-MAGA AI-generated avatar posts in January 2026 and identified at least 304 accounts spreading this type of content, some of which have since disappeared. Researchers at Purdue University’s Governance and Responsible AI Lab, known as GRAIL, found another dozen accounts spread across TikTok, Instagram, and Facebook. Eric Nelson, a special investigations analyst at digital threat mitigation firm Alethea, identified nine more accounts on YouTube.
None of the accounts reviewed by the newspaper were labeled as content generated by artificial intelligence. And perhaps most importantly: neither the New York Times nor the researchers consulted found similar networks operating in support of left-leaning candidates or causes. That detail matters because it shows that, at least at this scale and with this degree of coordination, the identified operation has a very specific political direction.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Who’s behind all of this remains a mystery. Experts say that determining whether these accounts are the product of a hired content farm, a foreign influence operation, an experiment, or something else entirely is extremely difficult. The consensus among researchers, however, is that creating this type of avatar is getting easier and cheaper by the day, especially for marketing firms and contractors that now specialize in developing and distributing AI avatars at scale.
Signs that give away the artificiality
Despite the visual sophistication, many of these accounts carry telltale signs that reveal their artificial origin to anyone paying attention. The same characters appear across multiple different accounts: a blonde with braids in a flowing dress on a farm at sunset, a woman in a purple blouse sitting in a wheelchair, a Black woman wearing a red MAGA cap and aviator sunglasses. Several of these profiles follow each other, forming interconnected clusters that share identical language, repeated images, similar profile photos, and even the same sound effects.
Most of the accounts claim to be from different American states, but they use broken English riddled with grammatical errors that expose the inconsistencies. Trump is called presidont in captions. Bios say things like sharing you the truth and ask viewers to follow me first if like my live. At least 13 accounts had nearly identical bios declaring Republican pride and asking Trump supporters to speak up. 🇺🇸
One particularly curious case is the avatar reposted by Trump himself, which has over 51,000 followers on TikTok. In its first two posts on January 15, the avatar spoke with a heavy foreign accent. By the next day, it had adopted an American accent. This kind of inconsistency is one of the classic signs of synthetic content that shifts rapidly as the operator adjusts the generation model’s parameters.
How these avatars actually work
The mechanics behind these accounts are more sophisticated than they appear at first glance. The artificial intelligence avatars are created with image and video synthesis tools that are already widely available on the market, some of them free or offered at affordable price points. They’re given digitally generated faces, cloned or synthesized voices, and even fictional life stories that make the profile more believable to someone quickly scrolling through their feed.
The result is a complete persona with a convincing profile photo, a crafted bio, and a posting style that mimics the behavior of real content creators — complete with stories, reels, and even comments answered in an automated fashion.
What makes this operation even more efficient is the volume. With hundreds of accounts operating in a coordinated manner, the same type of political message can reach different audiences on different platforms at the same time, creating that feeling that a topic is everywhere all at once. According to Zuhair Lakhani, co-founder of the AI advertising startup Doublespeed, backed by Andreessen Horowitz, each post featuring these pro-Trump avatars likely costs between $1 and $3 to generate. The company, which operates smartphone-powered bot farms and distributes hordes of synthetic influencers, says that a single person can now orchestrate what previously required a team of 30 creators and $40,000 — for just 10% of that cost.
Lakhani said Doublespeed turned down work with political campaigns, despite being approached by Democrats, Republicans, and foreign parties alike. According to him, the refusal came down to a moral compass. But he acknowledged that plenty of companies out there are taking those contracts, and that the money involved is substantial and very tempting. 🤖
The strategy behind mixed content
Not all of the identified accounts are exclusively devoted to politics. At least for now, some seem more focused on inflating engagement metrics, aggressively soliciting followers and comments. Others resemble romantic catfishing scams. Some sell hair removal creams or travel packages to China, filling their feeds with artificially generated content about entertainment, sports, religion, and other non-political subjects.
This mix of topics isn’t accidental. By alternating lightweight entertainment content with targeted political messaging, these accounts build an appearance of authenticity that makes detection harder for both platform algorithms and users themselves. A profile that talks about basketball, recipes, and Christian faith looks far more human than one that only posts political propaganda around the clock.
Kaylyn Jackson Schiff, co-director of GRAIL, noted that some accounts appear to be trying to reach specific audiences by making minimal adjustments to the avatars. The New York Times followed one account from early February, when it started posting as a dark-haired, brown-eyed avatar inside a car. Since then, the account published 37 videos. The avatar changed six times: the hair turned blonde, the eyes turned blue, and so on. These subtle tweaks suggest an A/B testing operation, where different versions of the same profile are tested to see which one drives more engagement with each type of audience.
Militarized avatars and the exploitation of conflict
These avatars frequently appear in military settings or dressed as immigration agents, as if trying to capitalize on recent conflicts. One Instagram account analyzed by the Washington Post featured an AI-generated blonde soldier avatar posing with Trump and other world leaders, and that account managed to amass more than one million followers.
During a single week in the spring of 2026, ten different accounts posted similar videos of women in military uniforms, all using the same artificial speech cadence to say that if it offended anyone, that was fine, just keep scrolling — before asking viewers to hit follow. Other posts show distorted American flags and other hallmarks of content generated by artificial intelligence.
Even though the quality of some of these videos borders on what’s now commonly called slop — that low-quality content mass-produced by AI — and some of the engagement may come from automated bot activity, researchers noticed something troubling in the comments on these posts: many real users believed the avatars were actual people.
The real impact on elections and public opinion
Saying this affects elections might sound like an overstatement to anyone who still associates political influence with rallies and televised debates, but the numbers tell a different story. When a single post from an AI avatar surpasses half a million views with zero synthetic content labeling, it enters the same information stream as verified news, expert opinions, and official statements. To the social media algorithm, engagement is engagement, regardless of who or what generated that content.
Andrew Yoon, a member of the technical team at CivAI, a nonprofit organization dedicated to educating the public about AI capabilities and consequences, put it bluntly: these avatars are trying to spread political messages and create the illusion of consensus. Flooding the zone with tons of videos appears to be a strategy designed to manufacture a false sense of majority opinion.
Researchers who study disinformation point out that the greater danger isn’t necessarily in convincing someone to change their vote directly, but in creating an atmosphere of alternative reality where certain claims appear widely accepted because they show up in many places at the same time. When you see the same story being told by ten different profiles across two different platforms on the same day, the human brain tends to interpret that as social consensus — even if all those profiles are controlled by the same operation. One expert described this approach as spray mode rather than precision, referring to the strategy of saturating the information environment instead of targeting specific individuals. 🧠
The episode in which President Trump reposted content from one of these avatars is emblematic precisely because it exposes a vulnerability that goes beyond the average voter. If a figure with access to advisors, communications teams, and verification infrastructure can be reached by a synthetic profile, the problem extends far beyond the general population’s digital literacy.
What the platforms are doing about it
TikTok, Instagram, Facebook, and YouTube all have policies requiring the identification of content generated by artificial intelligence, especially in sensitive contexts. The problem is that these policies largely depend on self-disclosure by whoever posts or on reports from other users.
TikTok stated that it conducted a careful review of the 304 accounts identified by the New York Times and found zero evidence of covert influence operations. The platform concluded that the accounts were spammers trying to generate engagement — something it classified as a regular and unfortunate occurrence on phones and social media. TikTok said it was in the process of removing those accounts.
Meta said it requires users, under threat of penalties, to disclose when photorealistic posts are created or edited using AI. However, it acknowledged that AI-generated content can be difficult to identify, especially as the technology evolves. YouTube stated it was reviewing the channels featuring political AI avatars and shutting down those that violated the platform’s policies against spam and deceptive practices.
After the reports came out, some platforms moved to remove the identified profiles, but digital security experts warn that this amounts to little more than a symbolic gesture given the scale of the problem. The infrastructure that created those 300 accounts can spin up 3,000 new ones in a matter of days, with different faces, different names, and slightly modified life stories designed to evade detection systems.
The Republican Party’s position
The Republican National Committee denied any involvement with the identified accounts. Zach Parkinson, the committee’s communications director, said Republican campaigns should use every tool at their disposal in their races, including artificial intelligence, but emphasized that the technology isn’t a silver bullet.
According to Parkinson, nothing replaces a winning message or a great candidate on the positive side, or real audio, video, and visual elements in an attack ad. In his view, authenticity is still king. That statement is noteworthy because it acknowledges the usefulness of AI as a campaign tool while simultaneously trying to distance itself from the specific practices revealed in the reporting.
Why this matters beyond American borders
It would be a comfortable mistake to think this is exclusively an American problem. Brazil, along with virtually any democracy that holds regular elections and has a population connected to social media, is equally exposed to this type of operation. The tools for creating AI avatars are global, accessible, and getting cheaper by the day, and the business model behind political disinformation campaigns doesn’t respect geographic borders. What the U.S. is dealing with right now is, in many ways, a testing ground for what could happen in any electoral process around the world in the coming years.
Digital literacy needs to evolve alongside the technology, and that includes teaching people to question not just the content they consume, but the very existence of whoever produces it. Checking whether a profile is real, verifying whether a face has been synthetically generated, being skeptical of accounts that pop up out of nowhere with thousands of followers and messaging perfectly aligned with one political side — these are skills that are becoming essential for navigating social media with a minimum of informational safety. It’s not paranoia; it’s a necessary adaptation to an environment that has changed dramatically over the past two years due to advances in generative artificial intelligence.
The picture the New York Times revealed in 2026 is a clear signal that the relationship between technology, politics, and truth has gotten far more complex — and that the tools we have today to deal with that complexity still fall well short of the challenge. The pro-Trump avatars that flooded the platforms are one chapter of this story, but they’re unlikely to be the last. The question is whether the responses — whether technological, legal, or educational — can keep pace before the damage to democracies becomes irreversible. 🌐
