Teachers in the United Kingdom have been advised to look out for specific words and phrases as potential indicators of artificial intelligence (AI) usage in student coursework, according to newly issued guidance from the Joint Council for Qualifications (JCQ). The recommendations, distributed at the start of the academic year, aim to help educators and examiners detect essays or assignments that may have been generated by AI tools such as ChatGPT.

The guidance highlights stylistic features and language patterns that could suggest AI involvement, including repeated use of transitional phrases like "firstly," "in conclusion," "however," and "therefore." Other suggested markers include overuse of words such as "delve" and "catalyst," frequent employment of em dashes and semicolons, and writing that exhibits a notably neutral or overly balanced tone, described as “safe and bland” compared to typical human expression. While the list does not present definitive evidence of AI use, JCQ officials said the indicators could assist teachers in identifying suspicious work.

JCQ, which represents eight of the UK's largest awarding organizations, maintains that students are prohibited from submitting AI-generated content as their own. The guidance reflects growing concerns about AI's impact on academic integrity amid advances in chatbot technology.

However, education specialists have criticized the list as ineffective and out of touch with current writing practices. Dr. Will Van Reyk, chief academic officer of the AI education platform Pedagogue and former deputy head at St Paul’s Girls’ School in London, described the approach as a “totally losing battle.” He argued that commonplace words flagged by the guidance are frequently used in human writing, and that students could easily instruct AI systems to avoid such terms. Instead, he proposed that coursework be redesigned to acknowledge and reward critical engagement with AI tools, with students required to document and assess the AI-generated content they use.

Iain Mansfield, head of education at the Policy Exchange think tank, also questioned the viability of coursework-based assessments in the AI era. He recommended shifting evaluation toward in-person exams and supervised assessments where AI use could be uniformly controlled. Mansfield emphasized that practical subjects might remain exceptions but asserted that most other assessments should adapt to ensure fairness.

A JCQ spokesperson emphasized the organization’s commitment to supporting schools in identifying and addressing malpractice, underlining that teacher judgment remains central to maintaining assessment integrity. An Ofqual representative added that presenting AI-generated work as one’s own constitutes cheating and carries serious consequences.

The guidance marks the first time a formal list of AI-related linguistic “warning signs” has been shared with UK educators, reflecting the increasing challenge artificial intelligence poses to traditional education assessments.