- One of the most pervasive forms of AI slop that Snopes debunks is content that aims to exploit American political divides for monetary gain.
- These stories are so common that Snopes has identified several recurring tropes, such as fake fights, exposés or insults between two prominent political figures. For example, in May, AI slop creators reaped money from advertisements attached to dozens of fake stories about U.S. President Donald Trump calling various prominent figures an “offender of Jesus.”
- This kind of low-quality AI content has gained less attention ahead of the 2026 midterms than AI content created by politicians, for example, or AI-generated propaganda.
- However, AI misinformation experts are still concerned about its potential impact on democracy — even if political scientists say there is no evidence slop impacts people’s voting behaviors.
- That is because democracy relies on an informed electorate. While falsehoods online are nothing new, AI has accelerated preexisting issues. Slop has flooded the online information environment, making it more difficult to access and be exposed to high-quality news.
- Social media platforms say they are trying to stem the flow of slop, but experts are skeptical their efforts are earnest.
A few months ago, Snopes readers asked us whether U.S. President Donald Trump had called WNBA star Caitlin Clark an “offender of Jesus.” Others wanted to know if he leveled that same insult at actor Kurt Russell, singer-songwriter Bob Dylan or pop star Taylor Swift.
We received so many queries about the “offender of Jesus” line that we published articles about it twice in two weeks. Still, requests flowed in. Recently, readers have also been asking whether Trump called rock musician Mick Jagger an “insult to Jesus.”
Despite Trump’s reputation for insults, it was not possible to find evidence of him ever publicly calling anyone an “offender of Jesus” — at least as of this writing.
So why have so many readers asked us about this particular insult?
Well, it’s part of the AI slop phenomenon. People around the world are making money by generating low-quality content, created using artificial intelligence tools, that seeks to invoke as many strong emotions as possible — often by exploiting American partisan divides.
As Snopes has previously reported, the AI slop business involves flooding social media algorithms, going for quantity instead of quality, until something goes viral. Popular posts bring in more money through advertising dollars or social media creator programs, which incentivizes slop creators to hop on trends based on what has worked before. That is how we end up with dozens of Facebook posts advertising different versions of the “offender of Jesus” insult. These stories are often based on real events, adding to the confusion — for example, Trump did actually call the beliefs of James Talarico, the Democratic U.S. Senate candidate for Texas, an “insult to Jesus.”
Ahead of the 2026 midterm elections, these low-quality slop stories have not received as much attention as AI deepfakes in political ads, for example, or Iran war disinformation generated using AI. Some political science experts we spoke to also said political AI slop does not necessarily change people’s voting behaviors or patterns.
But experts said it can still cheapen and distort our political discourse, entrench preexisting beliefs and move us further from the democratic ideal of an informed electorate that makes decisions based on fact, not AI fiction.
“Of course it’s harmful. We’re all swimming in this pea soup of AI slop,” Darren Linvill, a professor at Clemson University who studies digital disinformation, said in a video call. “It gives us this distorted view of reality. Anytime you have a distorted view of reality — an understanding of the world around you that’s not grounded in truth — that opens you up to all kinds of potential lies.”
The rise of political AI slop
Our team has seen major changes to the breadth and scope of political AI-generated content since Snopes first started covering AI slop in 2022.
A review of our fact checks tagged as examining AI-generated content found that in 2024, only 19 out of 86 total stories involved a public official or their family member (about 22% of our coverage).
In contrast, in 2025, we wrote at least 125 stories that involved political figures as defined above (out of 348), or roughly 36% of our stories about AI-generated content that year. That trend appears to be holding for 2026. From January to September, around 45% of our stories on AI content, 112 of 248, included public officials.
Some of these articles were fact checks that we rewrote after the same or a similar rumor spread again.
We also excluded articles about people who could be considered political figures but are or were not public officials — such as Erika Kirk and Charlie Kirk — and stories involving political topics, rather than individual people, in our figures above. If we had included them, those numbers would be even higher.
What we fact-check is largely driven by our readers’ inquiries — which may skew our data on AI slop — and our archives can only capture a tiny fraction of the AI-generated content spreading online. But we’re still seeing some clear themes to watch out for.
Perhaps the most easily recognizable trope is the fake fight, exposé or insult, like the “offender of Jesus” comment. We also often see stories about politicians or celebrities “storming off” an interview with television hosts such as Stephen Colbert and Rachel Maddow, who usually “reveal” fraud, corruption or other various crimes. (Maddow, incidentally, has shouted out Snopes for our work debunking AI slop about her.) Then there are the stories where one person “destroys” another or gives a “fiery” response that leaves their opponent, or sometimes an audience, “stunned.”
Another predictable trope is a story intended to embarrass or expose a political figure. These are often fake videos and photos of older politicians falling down, or images purportedly showing Trump and other lawmakers with the late sex offender Jeffrey Epstein.
On the other side, there are stories aiming to humanize politicians. These fall under the “glurge” trope, a term coined in the 1990s by an original member of the Snopes message board (before we became a news site). Snopes defined glurge in 2021 as a “sentimental or heart-rending story that undermines its own inspirational message by distorting — or ignoring — the facts.”
Examples include fake videos of the youngest Trump son, Barron Trump, performing on “America’s Got Talent,” or tech mogul Elon Musk supposedly saving the home of an elderly woman.
Broadly, most stories that did not fall within these tropes were still meant to provoke outrage and other strong emotions — such as a claim that Trump said he hoped Texas flood victims were immigrants, or a fake video thumbnail on YouTube showing migrants allegedly ambushing Marines.
In 2026, we saw fewer fights and exposés, but a new trope arrived — prominent people purportedly issuing a dire warning of some sort, usually about the state of American democracy and often issued late at night or early in the morning. See, for example, Colbert’s supposed “emergency monologue” at 3 a.m. or Pope Leo XIV allegedly beginning a livestream at that same time to inform his congregation he received a threat from a “powerful political figure.”
“One major problem with these AI distortions of reality is that they’re just so much more entertaining than reality, and so they spread,” Linvill said. “Even with [AI] Iran war content, the explosions are bigger. They’re more dramatic and the camera angle is always perfect. It’s just like we’re watching an action movie — and real war video isn’t always that nice to look at.”
‘Digital junk food’ on the democratic stage
According to Jason Miklian, who studies the “slop economy” — AI slop as a business — as a research professor at Norway’s University of Oslo, it is difficult to know exactly how harmful this deluge of slop is.
“AI slop inhabits this … nonsense realm,” Miklian said in a video call. “People aren’t getting fed misinformation or disinformation — it’s ‘noninformation.'”
Miklian still has serious concerns. In an October 2025 opinion piece for the Los Angeles Times, he and a colleague argued the slop economy “misleads voters, erodes trust in institutions and fuels polarization by amplifying sensational content.” They also warned that “people can become susceptible to extremism simply through prolonged exposure to slop.”
Speaking to Snopes, Miklian described political AI slop as “digital junk food,” meaning that if people consume slop content every once in a while, it’s not a huge deal, but the more slop people consume, the less factual information they are exposed to. People who cannot afford or simply are not interested in paywalled news content suffer more under the slop economy, he added.
“There’s very little research on it because it’s so new, but the little material we have seen is quite disturbing. Entire populations become less and less attached to a common reality,” Miklian said.
Falsehoods on the internet are nothing new. But Sejin Paik, a research professor at Georgetown University who studies perceptions of AI-generated political content, said in a video call that AI has driven the spread of low-quality information to an “uncontrollable speed.”
Paik and other experts said AI slop feeds into and accelerates existing concerns about online echo chambers — when algorithms expose people only to information that reinforces their preexisting views — and confirmation bias, or the concept that people are more likely to believe or favor information that supports their preexisting views because it makes them feel good.
“Political content like Nancy Pelosi or Donald Trump falling down the stairs — it’s created to connect with our prior beliefs, with our emotions about a thing,” Linvill said. “Anytime you have content that reinforces people’s beliefs, that’s going to entrench them in those beliefs, even if those beliefs aren’t necessarily what might be best for them or society.”
The ‘liar’s dividend’
Not everyone is as worried. Emily Thorson, a political scientist at Syracuse University who studies American politics and misinformation, said in a phone call that her research indicates political content is actually the “substantial minority of most people’s feeds.” Thorson said there is not much risk that AI slop could actually sway people’s votes because the people consuming it are usually already “locked into one side or the other.”
“There has long been a tendency for people to assume that folks are dumb and, ‘Well, they’re going to be hoodwinked by misinformation,'” Thorson said. “People aren’t that stupid. People approach what they see on social media with an enormous degree of skepticism.”
AI researchers, however, say the skepticism is leading to a different issue — people are now increasingly dismissing real, factual events as AI, something Snopes addressed when verifying an authentic video of Israeli Prime Minister Benjamin Netanyahu.
“We saw inklings of this in 2024,” Joshua Tucker, co-director of the New York University Center for Social Media, AI, and Politics, said in a video call. “It’s the liar’s dividend: Trump didn’t like the size of the crowds, so he just called it deepfakes.”
More recently, The Wall Street Journal reported in October that Kimberly Guilfoyle, Trump’s former ambassador to Greece, even tried to tell White House officials that texts suggesting she used her position for personal gain were AI-generated.
Tucker said he is far more concerned about people dismissing real events as fake, rather than the other way around. One of the risks, he said, is people might be more likely to choose sides based on partisan inclinations, rather than factual reasoning. The so-called “liar’s dividend,” Linvill added, feeds into concerns about confirmation bias and possible downstream effects in how voters make electoral decisions:
People will discount the truth in front of their eyes because it’s easy to discount the truth. That’s especially going to be a problem when the truth contradicts their prior beliefs, when the truth undermines what they believe and want to believe.
When you’re in a position where the truth makes you uncomfortable — if you have an opportunity to discount the truth, you might be more likely to take that opportunity.
A February 2024 study by researchers at the U.K.’s University of Cambridge indicates that claims of misinformation are more successful than other strategies politicians use to raise support among voters. That study found that it is easier to dismiss text as fake news rather than video. However, newer, emerging research indicates that as deepfake audio and videos circulate more frequently and become harder to distinguish from authentic media, people are increasingly likely to dismiss real audiovisual content as fake.
“You can decide whether you’re more worried about people being exposed to misinformation, or if you’re more worried about people hearing that everything is misinformation … and how elites are going to use that to try to escape responsibility,” Tucker said.
What’s the solution?
In an ideal world, regulatory measures would prevent AI slop from polluting the quality of information online, experts said.
However, government intervention on this issue is a thorny topic, especially with the United States’ strong First Amendment protections for free speech. The Trump administration hasn’t shown signs of wanting to slow down AI content. For example, in a Sept. 14 Truth Social post, Trump wrote, “AI taking over the World, destroying Humanity, and all other things bad, is a HOAX.” He has since signed an executive order renaming artificial intelligence as super intelligence, in which he celebrates the technology’s rapid progress.
As the backlash to AI grows among numerous groups, many social media platforms, including TikTok, have made at least some attempt to police themselves and crack down on the slew of slop, especially ahead of the midterms.
While Meta — which owns Facebook, Instagram and Threads — tries to label AI-generated content and has a general policy against spam, the company does not appear to have a specific policy explicitly aimed at tackling AI slop.
A Meta spokesperson said via email, “We will remove content, regardless of whether it is created by AI or a person, if it violates our policies.”
In contrast, YouTube says its monetization policies are aimed at discouraging low-quality AI content. In January, its CEO, Neal Mohan, listed “Managing AI slop” as one of the company’s top priorities for 2026.
YouTube considers most AI content that spreads false information subject to removal under the “spam, deceptive practices and scams” category. That category makes up the vast majority of channels that the video-sharing platform has removed for violating its community guidelines in recent months.
“YouTube’s business model relies on creating a high-quality environment that viewers, creators, and advertisers trust – and want to come back to,” a company spokesperson said via email.
It is also true that many of the stories we have cited above include social media posts that have since been taken down by various platforms for violating their guidelines.
Ultimately, many experts remain skeptical that social media platforms are able to — or even want to — police themselves.
“The platforms won’t stop it,” Linvill said. “This is working for the platforms, because people are sharing it. Because it’s what people want, whether they admit it or not. It’s engaging content. The explosions are better.”
In the absence of regulation, researchers (and fact-checkers) are scrambling to find other ways to inoculate the general population to AI slop, whether through media literacy education or AI-detection tools.
Paik, the Georgetown University professor, believes that those on the side of promoting high-quality information are “losing by a lot” — but that is not stopping her from working on a free, open-source AI-detection service, even if it is an uphill battle to keep pace with evolving AI technology.
“I don’t want to buy, ‘We lost the race and we’re just going to give up,'” she said.
