Out of 200 million papers reviewed, over six million have at least 80% AI writing present: Report

As the AI writing detector on Turnitin turns one, the company released the sobering statistic that shows how generative AI tools are shaping the way students and scholars carry out research

Turnitin’s AI writing detection feature was shortlisted by the 2024 Bett Awards in the “AI in education” category [File] | Photo Credit: REUTERS

Out of 200 million papers reviewed by AI writing detector and plagiarism tool Turnitin, over six million have at least 80% AI writing present, reported the company on Tuesday as its AI writing detection feature turns one.

Turnitin uses technology in order to assess digitally submitted papers and coursework to check them for plagiarism.

With the rise of generative AI tools such as OpenAI’s ChatGPT, the company in April 2023 launched its own AI writing detection feature.

Out of the roughly 200 million papers reviewed, around 22 million or roughly 11% had at least 20% AI writing present, per Turnitin.

“Turnitin’s dedicated AI engineering team began working on a detection solution more than two years prior to the launch of ChatGPT and was able to launch its new feature within months of the OpenAI’s generative AI application release. The tool integrates the AI writing report within the existing Turnitin workflow, providing educators with an overall percentage of the document that AI writing tools, like ChatGPT, may have generated,” said the company in its press release.

Turnitin’s AI writing detection feature was shortlisted by the 2024 Bett Awards in the “AI in education” category. It was also named Best in Show by Tech & Learning at ISTE 2023, per the company.

(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)

However, detecting AI writing is not an exact science and can be a controversial practice. ChatGPT-maker OpenAI itself last summer quietly disabled the AI writing classifier it had released to the public, citing its “low rate of accuracy”.

Several students have also accused Turnitin of delivering false positives, leading to penalties or disciplinary action at their educational institutions.

Turnitin addressed such allegations but defended its detection methods.

“Our sentence-level false positive rate is around 4%. This means that there is a 4% likelihood that a specific sentence highlighted as AI-written might be human-written. The incidence for this is more common in documents that contain a mix of human- and AI-written content, particularly in the transitions between human- and AI-written content,” said Annie Chechitelli, Chief Product Officer, Turnitin, in a blog post.