In May 2025, a post asking “[Am I the asshole] for telling my husband’s affair partner’s fiancé about their relationship?” quickly received 6,200 upvotes and more than 900 comments on Reddit. This popularity earned the post a spot on Reddit’s front page of trending posts. The problem? It was (very likely) written by artificial intelligence (AI).
The post contained some telltale signs of AI, such as using stock phrases (“[my husband’s] family is furious”) and excessive quotation marks, and sketching an unrealistic scenario designed to generate outrage rather than reflect a genuine dilemma.
While this post has since been removed by the forum’s moderators, Reddit users have repeatedly expressed their frustration with the proliferation of this kind of content.
High-engagement, AI-generated posts on Reddit are an example of what is known as “AI slop” – cheap, low-quality AI-generated content, created and shared by anyone from low-level influencers to coordinated political influence operations.
Estimates suggest that over half of longer English-language posts on LinkedIn are written by AI. In response to that report, Adam Walkiewicz, a director of product at LinkedIn, told Wired it has “robust defenses in place to proactively identify low-quality and exact or near-exact duplicate content. When we detect such content, we take action to ensure it is not broadly promoted.”
But AI-generated low-quality news sites are popping up all over the place, and AI images are also flooding social media platforms such as Facebook. You may have come across images like “shrimp Jesus” in your own feeds.
It costs almost nothing to make
AI-generated content is cheap. A report by the Nato StratCom Center of Excellence from 2023 found that for a mere €10 (about £8), you can buy tens of thousands of fake views and likes, and hundreds of AI-generated comments, on almost all major social media platforms.
While much of it is seemingly innocent entertainment, one study from 2024 found that about a quarter of all internet traffic is made up of “bad bots”. These bots, which seek to spread disinformation, scalp event tickets or steal personal data, are also becoming much better at masking as humans.
<!– Tag ID: zmescience_300x250_InContent_3
–>
In short, the world is dealing with the “enshittification” of the web: online services have become gradually worse over time as tech companies prioritise profits over user experience. AI-generated content is just one aspect of this.
From Reddit posts that enrage readers to tearjerking cat videos, this content is extremely attention-grabbing and thus lucrative for both slop-creators and platforms.
This is known as engagement bait – a tactic to get people to like, comment and share, regardless of the quality of the post. And you don’t need to seek out the content to be exposed to it.
One study explored how engagement bait, such as images of cute babies wrapped in cabbage, is recommended to social media users even when they do not follow any AI-slop pages or accounts. These pages, which often link to low-quality sources and promote real or made-up products, may be designed to boost their follower base in order to sell the account later for profit.
Meta (Facebook’s parent company) said in April that it is cracking down on “spammy” content that tries to “game the Facebook algorithm to increase views”, but did not specify AI-generated content. Meta has used its own AI-generated profiles on Facebook, but has since removed some of these accounts.
What the risks are
This may all have serious consequences for democracy and political communication. AI can cheaply and efficiently create misinformation about elections that is indiscernible from human-generated content. Ahead of the 2024 US presidential elections, researchers identified a large influence campaign designed to advocate for Republican issues and attack political adversaries.
And before you think it’s only Republicans doing it, think again: these bots are as biased as humans of all perspectives. A report by Rutgers University found that Americans on all sides of the political spectrum rely on bots to promote their preferred candidates.
Researchers aren’t innocent either: scientists at the University of Zurich were recently caught using AI-powered bots to post on Reddit as part of a research project on whether inauthentic comments can change people’s minds. But they failed to disclose that these comments were fake to Reddit moderators.
Reddit is now considering taking legal action against the university. The company’s chief legal officer said: “What this University of Zurich team did is deeply wrong on both a moral and legal level.”
Political operatives, including from authoritarian countries such as Russia, China and Iran, invest considerable sums in AI-driven operations to influence elections around the democratic world.
How effective these operations are is up for debate. One study found that Russia’s attempts to interfere in the 2016 US elections through social media were a dud, while another found it predicted polling figures for Trump. Regardless, these campaigns are becoming much more sophisticated and well-organised.
And even seemingly apolitical AI-generated content can have consequences. The sheer volume of it makes accessing real news and human-generated content difficult.
What’s to be done?
Malign AI content is proving to be extremely hard to spot by humans and computers alike. Computer scientists recently identified a bot network of about 1,100 fake X accounts posting machine-generated content (mostly about cryptocurrency) and interacting with each other through likes and retweets. Problematically, the Botometer (a tool they developed to detect bots) failed to identify these accounts as fake.
The use of AI is relatively easy to spot if you know what to look for, particularly when content is formulaic or unapologetically fake. But it’s much harder when it comes to short-form content (for example, Instagram comments) or high-quality fake images. And the technology used to create AI slop is quickly improving.
As close observers of AI trends and the spread of misinformation, we would love to end on a positive note and offer practical remedies to spot AI slop or reduce its potency. But in reality, many people are simply jumping ship.
Dissatisfied with the amount of AI slop, social media users are escaping traditional platforms and joining invite-only online communities. This may lead to further fracturing of our public sphere and exacerbate polarisation, as the communities we seek out are often comprised of like-minded individuals.
As this sorting intensifies, social media risks devolving into mindless entertainment, produced and consumed mostly by bots who interact with other bots while us humans spectate. Of course, platforms don’t want to lose users, but they might push as much AI slop as the public can tolerate.
Some potential technical solutions include labelling AI-generated content through improved bot detection and disclosure regulation, although it’s unclear how well warnings like these work in practice.
Some research also shows promise in helping people to better identify deepfakes, but research is in its early stages.
Overall, we are just starting to realise the scale of the problem. Soberingly, if humans drown in AI slop, so does AI: AI models trained on the “enshittified” internet are likely to produce garbage.
Jon Roozenbeek, Lecturer in Psychology, University of Cambridge; Sander van der Linden, Professor of Social Psychology in Society, University of Cambridge, and Yara Kyrychenko, PhD Candidate, Cambridge Social Decision-Making Lab, University of Cambridge
This article is republished from The Conversation under a Creative Commons license. Read the original article.
