Redditors Invented a Fake Word and an AI Site Published an Article Thinking It Was Real

Who's excited for "Glorbo?"

AI Copycat

Leave it to Redditors to pick a petty but oh-so-sweet fight with an AI-written website, tricking it into making an idiotic error.

Recently, dedicated perusers of the "World of Warcraft" subreddit realized that a shoddy gaming website called Zleague was ruthlessly scraping threads from the forum for content and regurgitating them in AI-generated news articles with little to no editing or oversight, and bylines credited to seemingly fake authors.

And thus, a trap was laid. On Thursday, a thread that now has over 2,000 upvotes was posted expressing excitement over the arrival of "Glorbo" to the game. Wouldn't it be great, the user suggested, if "some major bot operated news websites" published articles about this?

"I have to say, since they started hinting at it in Hearthstone in 1994, it was obvious that they would introduce Glorbo to World of Warcraft sooner or later," they added, joined by commenters in on the joke.

Gotcha!

It didn't take long for the website's AI to bite.

Later that day, an article titled "World of Warcraft Players Excited For Glorbo's Introduction" popped up, with some incredible "key takeaways."

"Players express excitement for Glorbo's arrival and its potential impact on the game," reads one.

"Some players have reservations about the mandatory item Klikclac" — another made up term — "and its effect on casual players," read another.

An entire paragraph is dedicated to "Glorbo's Impact on the Game," sharing that some players are worried that the developer could be "inconsiderate to non-Glorbo players," while others appreciate "Glorbo's backstory."

The original thread was soon updated: "WE F-CKING DID IT." And soon the AI-generated article was removed, though it remains forever enshrined in this archive.

Real Threat

As funny as it is, AI-run websites like these have very sobering implications.

The phenomenon has been trickling its way into the mainstream for a while. Earlier this year, the tech news site CNET quietly started publishing AI-written articles that later turned out to be rife with errors and even plagiarism. Publishers both large and small have followed suit, with executives clearly hoping to avoid paying human writers.

So if Zleague's "Glorbo" article is anything to go by, readers can expect more vacuously composed dreck like this in the future.

But we gotta hand it to the Redditors: they did stick to the man here, and maybe it's a trick that can be used to weed out impostors and copycats in the future. After all, dictionaries have pulled off the same trick for years.

More on AI: Google Secretly Showing Newspapers an AI-Powered News Generator