botorbuiltGuides

AI Content Farms: What They Are and How to Spot One

Updated October 15, 2026. 8 min read

In May 2023, the misinformation watchdog NewsGuard found 49 news sites that were written mostly by AI with no meaningful human oversight. By June 2026 its count had reached 3,749 sites in 16 languages. These are AI content farms, and you have almost certainly landed on one from a search result without noticing.

This guide explains what they are, why they keep multiplying, how they have changed, and what actually gives them away. We also share an uncomfortable finding from testing our own checker against them.

Check any website for signs of AI

Free, no sign-up. Every clue is explained.

What counts as an AI content farm

A content farm is a site that publishes huge volumes of cheap articles to capture search traffic and sell ad impressions. Content farms existed long before AI. What changed is the cost: a single operator with a language model can now publish hundreds of articles a day.

NewsGuard's tracking center uses a strict four-part test. A site qualifies only if a substantial share of its content is clearly AI-produced, it is published without significant human oversight, it is designed to look like a normal human-run news outlet, and it doesn't disclose that the content is AI-generated. That last condition matters: a site that openly says its summaries are machine-written is not what we are talking about here.

Why they exist: the money

The business model is simple. Articles attract search visitors, visitors see ads, ads pay per impression. Most of those ads are placed automatically by ad networks, which is how, as MIT Technology Review reported, ads from major brands ended up funding junk sites without the brands knowing. Some farms add affiliate links or product reviews on top.

When producing an article costs a fraction of a cent, even tiny ad revenue per page is profitable at scale. That is the whole engine behind the growth from 49 sites to several thousand.

How they changed between 2023 and 2026

2023: new domains and obvious mistakes

The first wave was easy to spot once you knew what to look for. Sites were mostly on freshly registered domains with generic names. Some published articles that still contained chatbot error and refusal messages, which is how NewsGuard caught several of them in its first report. One site published a fabricated celebrity death story that included the text of a chatbot declining the request.

2025 and later: old domains and cleaner text

The current generation is harder. Operators now buy expired domains that once belonged to real newspapers, college radio stations or small businesses, because those domains already have search-engine trust and backlinks. Nieman Lab documented a wave of these in 2025, including a site impersonating a college radio station and a company claiming to have built about 1,500 sites with an AI tool. The text has also improved. Leftover chatbot messages are rarer, and the most famous "AI words" have largely disappeared.

How to spot one

No single sign is proof. Look for several at once.

The domain used to be something else

Open the domain in the Wayback Machine. If a "news" site was a bakery, a band's homepage or a local paper that closed years ago, and the current content started recently, you are looking at a repurposed domain. This is the strongest single check for modern farms.

Topics that sprawl in every direction

Celebrity gossip next to mortgage advice next to gardening tips next to crypto, all under a name that sounds like a local newspaper. Real outlets have a beat. Farms chase whatever is searched.

Authors with no life outside the site

Click the bylines. Farm authors rarely exist anywhere else, often have generic headshots, and publish an implausible number of articles per day across unrelated topics.

News that is really someone else's news

Many farms rewrite stories from established outlets within hours. Search a distinctive sentence or a quote from the article. If the same facts appeared elsewhere first, with no credit and no added reporting, the article is likely a machine rewrite.

Volume that no newsroom could produce

Check /sitemap.xml or the archive pages. Hundreds of articles a week from a site with no visible staff is the clearest numerical sign. Be careful with dates in sitemaps, though: some platforms rewrite all of them on every rebuild, so look at the article pages themselves.

Leftovers and stale facts

Search the site for phrases like "as an AI language model" or "I cannot fulfill". It still happens. Also watch for old events reported as new and confident details you can't find in any other source.

An ad layout that comes first

Ads between every paragraph, autoplay video, and a page that is mostly ad slots with a thin strip of text in the middle. Not proof on its own, since plenty of human sites over-monetize, but it fits the model.

What we found when we tested our own checker

We built Bot or Built to detect AI-made sites automatically, and we tested it on sites that NewsGuard, MIT Technology Review and Nieman Lab had named as AI-generated, alongside 50 well-known human-written sites. Two findings are worth sharing.

First, many of the farms named in 2023 no longer exist. Their domains have expired or now redirect elsewhere, which is typical of a business that treats sites as disposable.

Second, and more humbling: our checker missed most of the farms that are still online. Style signals such as AI-favourite words, em dashes and sentence rhythm looked the same on these farms as on human blogs. Their domains were often old, which makes them look established. They run on ordinary WordPress, so there is no builder fingerprint to find. The checks that do catch them, like comparing a domain's current content with its archived history or tracing where a story first appeared, still need a person.

We would rather say this plainly than pretend otherwise. Automated tools are good at spotting sites generated by AI builders. For modern content farms, the manual checks above are still your best tool.

What to do if you find one

Check any website for signs of AI

Free, no sign-up. Every clue is explained.

More guides