Pangram Releases New AI Writing Detection Model

Pangram, the most accurate and most reliable tool for detecting text created by AI, has released a new, better detection model. The upgrade improves on Pangram’s already unmatched accuracy and includes a unique feature that will aid educators when assessing student-submitted writing.

The new model, already released to customers as part of regular updates and improvements, has been re-engineered and retrained, and is proving even more accurate and successful in identifying AI-generated text from the newest models of AI providers such as OpenAI, Anthropic, and Gemini. Pangram’s new model is also, it seems, able to detect AI-created text from models that are awaiting public release.

“Our detection technology was already the best because we built it differently,” said Max Spero, Pangram’s CEO. “Pangram does not rely on perplexity and burstiness like other AI classifiers. Pangram is trained using ‘synthetic mirrors’ of the hardest documents to classify and then it is retrained over and over again. That makes it adaptable to new models and doesn’t require significant re-engineering or time-intensive retraining to remain relevant.”

The new detection capabilities also are successful at detecting “humanized” text – text that has been put through an automated paraphrasing engine. Humanizing text is a common tactic used by students and others who may wish to hide or contaminate the origins of text. Several humanizing products explicitly advertise their ability to beat AI detection. Although, in internal testing, even these humanizers fail to bypass Pangram.

“The days of being able to ask ChatGPT to do your assignments, then wash it with Grammarly or QuillBot, and expecting to get away with it – those days are coming to an end. If they’re not already over,” said Spero. “With Pangram, teachers are going to spot the use of humanizers with precision and regularity,” he said.

In addition to the significant upgrades, Pangram has released a new feature that will assist educators in identifying text created by AI, and, more importantly, in making sound decisions about how to address it. Starting immediately in educator dashboards, Pangram results will now return a result of “mixed” for submitted text, in addition to the classifications of “human” or “AI.”

This “mixed” designation will be applied to text that is likely to be partly human composed, mixed with portions that are likely produced by AI. Further, the “mixed” designation will break down the percentages of each – human versus the percentage that is detected as AI. Finally, in a “mixed” result, educators will be able to see the specific segment of the text in each category – human or AI.

Insight into mixed text is important because many AI users, especially students, generate AI answers or solutions but then edit the results themselves, hoping to either fool an AI detector or add some of their own voice, or both. 

Subscribe to edCircuit to stay up to date on all of our shows, podcasts, news, and thought leadership articles.