Nonprofits Launch Free AI Writing Checker To Prove If Works Are Bot-Composed

Sooner or later, the news reports you read could be all AI-generated. It’s already happening.

As we speak, the European Union is ironing out new laws to push for transparency in the unmoderated world of artificial intelligence generators, a new curveball in an ongoing fight against misinformation and disinformation. While those remain in the works, two literacy nonprofits have banded together to develop a tool to verify if something was written by programs like ChatGPT.

AI Writing Check, created by Quill and CommonLit, is open-source and completely free to use. Work for it began in December, around the time of ChatGPT’s launch, and when the threat of machines being utilized to help students cheat became very real. The AI model joins an array of checkers built hurriedly to combat the rise of bot-written assignments. Another version, GPTZero, is permanently free for users who opt for its ‘Classic’ mode. Turnitin, a platform integrated into school portals for students to submit their assignments, is also working on an AI detector to fortify its own plagiarism checker.

With AI Writing Check, educators and others faced with suspect bodies of text can simply copy and paste writing in a box and find out, in just a few seconds, if it may have been artificially composed by ChatGPT. You don’t need to sign up to use this feature, which is funded through philanthropic efforts to remain free for all.

The AI plagiarism checker was developed, first and foremost, to preserve the integrity of academic work, CommonLit CEO Michelle Brown tells Fast Company. It’s essential that students continue to do the “heavy lifting,” as the generations before have done, and learn that shortcuts aren’t the way to go for long-term success. AI Writing Check is free so under-resourced schools can still access such tools, adds Quill founder Peter Gault.

In a bid to stem a trend of “AIgiarists,” ChatGPT revealed that it is secretly  “watermarking” works so researchers with the key to decode them can determine if a body of text was crafted by a human or AI. Although the results may read as organic to many, the choice of grammar and vocabulary is based on a sequence and makes the AI-written work “statistically predictable” to the trained eye.

ChatGPT’s potential to advance or hold learning back can’t be overemphasized. The tool, which is free to use and can handle unlimited text prompts, gained a million users within five days of its launch in November. As noted in a graphic shared by ChatGPT on LinkedIn, it took Netflix around 3.5 years to hit the milestone, Facebook 10 months, and Instagram two months to reach the one-million mark. Many schools have now banned the use of ChatGPT, but there’s no stopping some students from defying these rules.

All told, AI Writing Check’s creators caution that it’s only 80% to 90% accurate, as determined by a scan of 15,000 essays.

For now, one of the smartest ways to catch a cheater unawares is by quizzing them on the spot, says philosophy professor Darren Hick, who made headlines last month after discovering that his student had fed a class assignment through ChatGPT.

Head here to try AI Writing Check for yourself.

[via Fast Company and CommonLit, images via various sources]