Wake student fights back after AI detector flags her work
Eleanor Canina received a failing grade on an English assignment about "Romeo and Juliet" with a note saying "evidence of AI, Please redo." The Green Hope High School freshman had written the essay herself. Now she and her mother are pushing back against the school's reliance on AI detection tools that research shows produce false positives.
The teacher ran Canina's assignment through three AI assessment tools. They returned likelihoods of 62%, 75%, and 87% for AI generation or significant AI assistance. When Canina's mother, Stacy De Coster, suggested the teacher compare the assignment to her daughter's past work or check the Google Doc history, the teacher declined.
De Coster, a sociology professor at N.C. State, said the approach violated her own institution's standards. North Carolina's Department of Public Instruction explicitly warns schools against relying on AI detectors as a sole measure of academic integrity.
What the guidelines say
DPI's guidance states: "AI detectors have proven not to be dependable, therefore they should never be used as the only factor when determining if a student 'cheated.'" The guidelines note that detectors frequently flag work by non-native English speakers and students with advanced vocabularies.
Wake County said it does not require AI detection tools and instead encourages teachers to review students' writing processes and work history. The district's statement acknowledged that AI assessment remains "a new and rapidly evolving area."
A New York judge ruled in February in favor of an Adelphi University student accused of plagiarism based solely on an AI detector result, signaling broader legal concerns about the tools' reliability.
The broader problem
Canina is an avid reader with a developed vocabulary. De Coster suspects her daughter was flagged because she used terms like "the titular character." She worries that strong writing now carries risk.
"Now she's writing under the specter of the possibility of false accusations," De Coster said. "So I don't know if I should tell her, 'dim your light, write at a lower level, just so that you don't get flagged.'"
Canina created an online petition calling for protections against false AI detection results at Green Hope. One commenter reported being accused of using AI because the tool flagged "intense vocabulary and few grammatical errors."
The student's position
Canina is critical of actual AI use in education, citing both environmental and ethical concerns. She said students who use AI to avoid work "isn't going to help anyone in the long run."
But she separated that concern from the detection problem. "I understand and acknowledge your concern regarding the limitations and variability of AI detection tools," the teacher wrote to De Coster. Yet he offered only an alternative assignment, not a resolution to the underlying issue.
Green Hope offered to have another English teacher grade the original assignment. The family refused, saying the solution doesn't address why teachers are relying on unreliable tools.
Canina said she won't change how she writes. "I want the system to change, not me to have to change," she said.
Learn more about AI for Education and best practices for AI for Teachers.
Your membership also unlocks:
