AI Detectors: Why They’re Unreliable (and what teachers can do instead)
As the use of AI increased, many educators turned to AI detection tools as a way to crack down on plagiarism, but now realise that in practice, they are far less reliable than expected.
Let’s start with why AI detectors are appealing:
They typically promise to:
identify AI-generated text
provide a confidence or probability score
help teachers determine whether a learner’s work is “authentic”
So, what’s the problem?
AI detectors do not definitively identify AI-written work
They frequently produce false positives
Often struggle with short answers, edited text, or mixed human–AI writing
High-performing students, neurodivergent learners, and second-language writers are often disproportionately flagged.
A detector score is not hard proof of plagiarism.
Why these tools are becoming less reliable:
Even when detectors improve, accuracy is unstable because:
AI models update rapidly and are constantly improving
Writing styles evolve over time
Students are finding ways around detection – e.g., using humanisers, rewriters, or paraphrasers
Detectors are not a long-term solution.
The risk for schools:
Over-reliance on AI detectors can lead to:
inconsistent application of academic integrity processes
challenges when decisions need to be defended
Though they may prompt a closer review of work, AI detectors should never replace professional judgement or be used as the sole basis for academic decisions.
Heavy reliance on detection tools can create a damaging culture of fear and mistrust.
When students believe they may be accused of misuse of AI, it can:
· discourage risk-taking and genuine learning
· damage student confidence and wellbeing
· shift the relationship from learning partnership to surveillance
What’s better than detection?
More robust approaches include:
assessment design that values process, not just product
reflective or explanatory components
personalised or contextualised tasks
clear expectations about acceptable AI useThese approaches reduce reliance on detection altogether.
AI detection is one of the most common topics we are asked about. TAKE AI focuses on practical alternatives to AI detection that support fairness, confidence, and academic integrity.