There Are Now More Fake News Sites Than Real Ones

There are now more fake local news websites in the United States than real ones. That’s not a projection. That’s the count: as of early 2026, more than 1,200 pink slime sites have been identified and cataloged by the Tow Center for Digital Journalism at Columbia University, outnumbering the remaining legitimate daily newspaper websites.

The natural response is: that sounds like a fake news problem. It isn’t. It’s worse. It’s an accuracy problem. These sites don’t publish false information. They publish true information selected and framed to serve political agendas. And that distinction is the entire game.

This is an accuracy trap: content that passes every fact-check while being structurally designed to manipulate you. It is the hardest form of manipulation to detect because the facts are real. The manipulation is in what gets covered, what gets ignored, and how the framing steers your conclusions.

What Pink Slime Sites Are

The term comes from the food industry. Pink slime is mechanically processed meat that is technically edible but bears little resemblance to what you think of as food. Pink slime news is mechanically generated content that is technically accurate but bears little resemblance to what you think of as journalism.

Here is how they work. A politically funded operation registers dozens or hundreds of domain names designed to look like local news outlets. Names like “The Kalamazoo Times” or “The Iowa Standard” or “The Mecklenburg Herald.” Each site publishes a mix of genuine local content (pulled from press releases, government filings, police blotters) and partisan framing pieces. No bylines. No editorial staff. No corrections page. No newsroom.

The content is increasingly generated by AI. Not the crude, error-filled AI content you might expect. Clean, professional-sounding copy that reads like any competent regional newspaper. Because the underlying facts are accurate, the sites pass basic credibility checks. They show up in Google News. They get shared on Facebook. They look, to a casual reader, like the local paper.

They are not the local paper.

Why Bias Ratings Miss This

This is where it gets uncomfortable for the existing solutions. AllSides rates media outlets on a Left-to-Right spectrum. Ground News labels coverage by political lean. These are useful tools. I use them.

But pink slime sites break the model. An article about a local school board meeting that reports accurate vote counts, accurate budget figures, and accurate quotes from board members does not register as biased. The manipulation is not in what the article says. It is in the pattern of what gets published and what doesn’t.

A pink slime network covering education might publish every story about budget overruns and teacher complaints while ignoring test score improvements and program expansions. Each individual article is factually accurate. The aggregate coverage produces a misleading picture. No bias rating catches this because bias ratings evaluate individual articles, not editorial patterns.

Our taxonomy at ntrl classifies this as Category 5 (Structural and Editorial) and Category 6 (Incentive and Meta) manipulation. The manipulation lives in the selection, the emphasis, and the hidden funding sources. It is invisible at the article level. It only becomes visible when you zoom out and ask: who is funding this, and what picture does the full body of coverage paint?

The 2026 Midterm Accelerant

Pink slime sites are not new. The first wave appeared before the 2020 elections. But the scale has changed.

In 2019, researchers identified roughly 450 pink slime sites according to a Tow Center report for the Columbia Journalism Review. Today: more than 1,200 and counting. The growth correlates with two developments: generative AI made content production essentially free, and the continued collapse of local news left more information vacuums to fill.

That second factor matters more than the first. When a real local newspaper closes, the community doesn’t stop needing local information. They still want to know about school boards, city council votes, zoning disputes, police activity. Pink slime sites move into that vacuum. They provide the information people want, wrapped in the framing their funders want.

With the 2026 midterms approaching, researchers at Yale’s Institution for Social and Policy Studies, Columbia’s Tow Center for Digital Journalism, and the Reuters Institute for the Study of Journalism at Oxford are tracking what amounts to the fastest expansion of coordinated partisan media infrastructure in American history. And most people encountering these sites have no idea they’re not reading genuine local journalism.

What This Means for You

If you follow news from multiple sources (and you’re reading this blog, so you probably do), pink slime content has almost certainly appeared in your feed. It looks like a local news article. It reads like a local news article. The facts in it check out.

The tell is in what’s missing. No byline. No editor listed. No corrections history. No physical address for the newsroom. And most importantly, no coverage of stories that would complicate the site’s political thesis.

This is why “just check the facts” is insufficient advice for navigating modern media. The facts can be fine. The manipulation can still be real. Checking individual claims doesn’t catch editorial manipulation because the manipulation is not in the claims. It’s in the selection.

What We’re Building For

At ntrl, we remove manipulative language from news articles. That’s our core product. But building the neutralization pipeline taught me something I didn’t fully appreciate when I started: linguistic manipulation is only one layer of the problem.

The language in a pink slime article might be perfectly clean. No loaded words, no urgency inflation, no emotional appeals. Our neutralizer might scan it and find nothing to change. And the article would still be part of a manipulative system.

This is the frontier. The manipulation that can’t be caught at the sentence level. The manipulation that lives in patterns, in funding, in what gets published and what gets buried. We’re not there yet. Right now, we remove manipulative language. But the taxonomy we’ve built, the framework we use to classify manipulation into six categories, was designed from the start to accommodate structural and systemic manipulation, not just linguistic manipulation.

The hardest form of manipulation is the kind where every individual piece of content is true. Pink slime sites are proof that accuracy, by itself, is not enough. You also need honest selection. Honest framing. Honest funding. And transparency about all three.

That’s what we’re working toward. One layer at a time, starting with the language.