Researchers from Northwestern University and the University of Tübingen have developed a method to distinguish AI-generated text from human-written content. Their study, which analyzed 14 million academic article abstracts from 2010 to 2024, revealed a sudden increase in the usage of certain style words following the emergence of large language models (LLMs). The researchers estimate that at least 10% of 2024 abstracts were processed using LLMs. The study identified nearly 300 words that became abruptly more popular, including “delves,” “showcasing,” and “crucial.” Unlike previous word usage spikes, which typically involved content-specific nouns related to current events, this shift focused on adjectives and verbs. The researchers argue that this unprecedented change in vocabulary can serve as a marker for detecting AI-generated text, offering a potential solution to the growing challenge of identifying AI-authored content in various fields.
AI Text Detection – New Study Reveals Linguistic Clues
Study finds AI-generated text uses certain style words more frequently, offering a potential method for detection.
1–2 minutes
TOP STORIES
Courts are now imposing serious penalties on attorneys for using AI hallucinations in legal filings …
Major tech companies are converging on agentic AI platforms for the physical world …
Devin Kim claims he was fired for raising AI safety concerns at xAI …
Nvidia’s acquisition of Kumo AI aims to enhance its predictive analytics capabilities …
Meta partners with Reliance to build a major AI data center in India …
The race for physical AI technology highlights the contrasting strategies of Nvidia and Tesla in the U.S.-China tech competition …
