The Continued Rise of ‘Parasitic Plagiarism Merchants’

Rewrite/edit

In late April, the Australian news show Media Watch ran a story about a website, League Initiative, ripping off local sports coverage and hiding the plagiarism by having AI systems rewrite the content.

Patrick Woods, a reporter at the Townsville Bulletin, initially blew the whistle after noticing several of his stories on the site. He referred to this type of site as “parasitic plagiarism merchants” and said that they were “becoming all too common.”

The (re)writing on League Initiative was inferior, to the point of comedy. In one comical highlight, the site’s AI rewrote the line “We believe he’s only just scratched the surface of his potential” to “He’s nudged the iceberg, but there’s so much more to come.”

This could have been the end of the story. With the site called out and few clues as to who was operating it, it could have been a dead end. But it wasn’t.

On May 13th, ABC News ran a follow-up article. In that article, they first revealed that League Initiative was just one site of many, making up a network that plagiarized over 12,000 articles.

ABC then examined archived copies of the site and found references to an author named James Raptis. At the time, Raptis was senior legal counsel at Australian Community Media, a major media conglomerate.

When ABC News contacted Raptis about the sites, they quickly went offline. In a statement to ABC News, he denied ever writing for the sites but admitted to hosting them. He said he was unaware of the plagiarism despite helping set up the sites.

Yesterday, Australian Community Media announced that Raptis had resigned from the company, effective immediately.

It’s the end of a bizarre story, but a story that is becoming much more common on the internet.

The Continued Rise of AI Spam

In June 2023, I discussed the rise of AI spam. At the time, Newsguard, a company that vets news sources, reported 277 Unreliable AI-Generated News and Information Websites (UAINs). Today, that number is 840.

However, that is likely just the tip of the iceberg (or the nub of the glacier). Newsguard doesn’t track or monitor every news site. There are likely many, many times more UAINs out there.

The reason is simple: These sites are incredibly easy to set up and equally difficult to detect and close.

One needs only create a bare-bones website, pull content from various sources and have an AI system rewrite the articles. In theory, this creates “original” content for search engines, which makes it ripe for advertising.

To be clear, we’ve been here before. In the late 2000s, the threat wasn’t AI but article spinning. Such spinners could “rewrite” content automatically, at least enough to fool search engines, and they enabled a similar wave of spam sites.

However, article spinning crashed in early 2011 following a series of Google updates. Those updates improved the detection of such sites and demoted them, making them no longer viable.

However, with AI, detection is much more difficult. However, even if Google could reliably detect AI writing, it’s unclear how strong a stance it would take. The company has already softened it’s position on AI content and has invested heavily in its own AI systems, including within search results.

Though Google has made some updates to reduce AI garbage, the effects are dubious. This enables sites like League Initiative to thrive despite their parasitic nature.

Unfortunately, they may be all but impossible to stop.

The Problem with Prevention

Right now, the problem with stopping sites like League Initiative is simple. While humans can often spot AI-generated works, there’s no reliable automated way to do so.

Though humans may laugh at lines such as “He’s nudged the iceberg,” computers have trouble spotting it for the same reasons AI systems write it: It’s technically correct.

While we recognize it as an awkward turn of a common expression, it’s only wrong when you understand the context and deeper meaning. Even now, computers don’t understand the actual meaning of what they create.

This is why AI systems are prone to hallucination, why image generators struggle with hands and why you get phrases like “He nudged the iceberg.” AI is trying to guess the next word or pixel without a good understanding of the actual meaning.

Unfortunately, that works in both directions. When trying to detect AI, algorithms can’t spot the issues humans can. That is unlikely to change.

That’s not to say that the problem is impossible or can never be solved. It simply means that current approaches don’t work.

Because of that, sites like League Initiative aren’t just going to become more common. They’re going to become the norm.

Bottom Line

The story of League Initiative and its network of sites highlights the challenges of stopping AI spam. It took a human journalist, Patrick Woods, to spot the apparent similarities. It then took an intense human-driven investigation to find the rest of the network and a name connected with it.

This wasn’t a site or a network weeded out of search engines. It was a network publishing thousands of AI-rewritten articles that were making revenue from advertising using traffic they got from search results.

While there are steps we can take now, such as advertising networks being more cautious about what sites they allow, those steps are limited without an automated way to detect AI writing.

AI spam is easily the most predictable result of the generative AI boom. Once spammers had a simple way to produce mountains of “original” content, it was destined to create a new wave of spam sites.

As Jon Gillham, the founder of Orgiinality.ai, put it, “Not all AI content is spam, but I think right now all spam is AI content.” While that probably isn’t literally true, it’s easy to see why it feels that way.

Sadly, that’s something that isn’t likely to change any time in the near future.

Header Image: Still from Original Media Watch Report

Want to Reuse or Republish this Content?

If you want to feature this article in your site, classroom or elsewhere, just let us know! We usually grant permission within 24 hours.

Click Here to Get Permission for Free