AI slop is taking over the internet. And it's here to stay

By Christine Ji and Britney Nguyen

AI tools are more powerful than ever, but they're creating an internet flooded with low-quality content

Innovations in AI image and video generators have achieved unprecedented fidelity this year, largely conquering previous hurdles like rendering text and human anatomy.

In just three years, generative artificial intelligence has gone from a fringe Silicon Valley novelty to a household name.

The technology's widespread adoption has come at a steep price - and that's not only for data centers and energy bills. While public sentiment isn't as easily measurable as capital expenditures, the fact that Merriam Webster crowned "slop" the word of the year for 2025 suggests that AI has fallen on the wrong side of the zeitgeist.

To see why AI has developed a reputation associated with unappetizing pig feed, you just have to start scrolling. AI-generated articles, photos, videos, e-commerce listings and Spotify artists are all examples of slop: low-quality, low-effort media mass produced by AI tools, proliferating on the internet.

To Joel Lewenstein, head of product design at Anthropic, AI slop is "generic content, undifferentiated content - something that optimizes for volume and passive consumption over everything else."

Sometimes there are telltale indicators, such as highly repetitive phrases (including the classic "it's not just X, it's Y") or images with a signature yellow tint. However, as AI models improve, spotting the difference is becoming increasingly difficult.

"It's like AI junk food," Lewenstein added. "It fills up space, but it doesn't actually add any substance."

The rise of slop

AI seems to have found itself in a precarious paradox thus far: the more capable the tools become, the more intense the tsunami of slop.

According to research firm Graphite, the amount of AI-generated articles on the internet surpassed human-written ones in November 2024. This year saw the rise of several groundbreaking image and video models such as OpenAI's Sora and ChatGPT Images, as well as Google (GOOGL) (GOOG) Gemini's Nano Banana. And in 2026, Meta Platforms (META) is rumored to be releasing a new image and video model dubbed "Mango."

Read: Google's winning streak continues as viral Nano Banana gets an upgrade

Advancements in visual media generation have been especially impressive. The technology has largely conquered previous pain points in rendering text and fingers, and has gained widespread popularity as a result. Nano Banana sent Gemini to the top of Apple's (AAPL) U.S. App Store in September this year, for example.

People who aren't particularly skilled at writing or art now have access to tools that can help them, Paul Chan, founder of AI operating system Decidr, told MarketWatch. But to many people, the ability to generate massive amounts of media with just a few keystrokes has commoditized the process of creation itself and could be on track to destroy the livelihoods of millions of people.

See more: Believe the hype: AI is coming for your job. Plan now.

These tools also make it easier to spread misinformation and commit fraud, eroding social trust in digital ecosystems, Alon Yamin, CEO of AI-detector platform Copyleaks, told MarketWatch. When AI can generate convincing photos and text, "'seeing is believing' is no longer the case," he said.

All of these factors will likely lead to the amount of AI slop around us increasing, both Yamin and Chan believe.

Ziwei Cong, a marketing professor at Georgetown, sees AI slop bifurcating the internet. As AI-generated content becomes even cheaper and more abundant, she expects human-created content to retreat behind expensive paywalls.

There will still be demand for human-led craft, "but in the age of AI, the cost of human creation becomes so high" that it will be relegated to a very small, high-end market, Cong told MarketWatch.

A post-slop world

Beyond flooding the internet with commoditized content, techno-optimists worry that AI slop has stigmatized the use of AI for legitimate productive purposes.

AI has the potential to speed up drug discovery, personalize education and power robots - breakthroughs that could dramatically improve lives and add trillions in economic value.

If people are exposed to a high amount of what they deem AI slop, "that's what they will associate with AI, and that will be a detriment and a disservice to all of the positive outcomes that AI can generate," Anthropic's Lewenstein told MarketWatch.

The reputational damage seems apparent: a recent survey conducted by Anthropic found that while 86% of professionals felt like AI was additive to their workplace productivity, 69% mentioned feeling social stigma surrounding its use.

Read: How can Anthropic stand out in the AI wars? I went to a Greenwich Village pop-up to find out.

This pushback and negativity isn't uncommon in the technology lifecycle, however. After all, AI-generated videos of anthropomorphic animals and factually incorrect chatbot responses are not what AI is about in the long-run, Chan said. Rather, they are a symptom of being in the early stages of the innovation process.

Chan told MarketWatch that when he was coming up in the tech world more than two decades ago, there was a similar negative reaction to the emergence of the internet.

"People started to get sick of the word 'internet' or anything internet-related, and the stock market was very bullish as well," Chan said, drawing parallels between then and now. At the time, some websites that were being born out of the dot-com boom were considered slop, he said.

"Publishing initially was only for the super tech-enabled because it was really hard to make a webpage," Chan added. As the internet made access to publishing widespread, more people could publish whatever they wanted on web blogs, forums and eventually social media.

Chan and Lewenstein are optimistic that the benefits of AI will prevail over the sentiment of slop. In Chan's view, as more tangible use cases move into the mainstream, they will displace the high-volume, low-effort junk that defines the present moment. Realizing the technology's full economic value will take time, he noted, mirroring the decades-long arc of the internet's global adoption.

Lewenstein believes AI's true promise lies in augmenting, not automating, human capabilities. He encourages users to engage with the technology as a dynamic co-pilot for tasks ranging from debugging code to planning a vacation.

However, the proliferation of slop presents a lasting challenge as AI capabilities blur the line between human and AI creations.

"There are a lot of important aspects we need to think about if we want to make sure that we have safe AI usage in the future," Yamin said. "Understanding what's created by AI, what's real, and being able to differentiate between the two comes very important in mitigating these new risks."

-Christine Ji -Britney Nguyen

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

12-29-25 0836ET