How to check if a paper was written by AI
To check if a paper was written by AI, look at the writing process rather than the finished text. Was the paper typed gradually with evidence of drafting and revision, or did large sections arrive via paste with minimal editing? Process evidence gives you concrete, defensible facts; running the prose through an AI text detector gives you a probability score that can be challenged.
Why text detectors struggle with academic papers
Academic papers often use formal, structured language that can score as 'AI-like' even when written by a human. Non-native academic writers are especially vulnerable to false positives. Several universities have stopped using text-based AI detectors in disciplinary proceedings for this reason.
The case for process-based paper checking
Process signals (a 3,000-word paper submitted in 11 minutes; 94% of the text arriving in two paste events) are concrete facts that are hard to dispute. They don't depend on what the prose sounds like, they're the same regardless of AI model, and they're fair to every student.
Practical steps for teachers and institutions
Collect coursework through a process-aware submission tool so the writing timeline is captured. Review behavioural signals before reading the final text. Use flagged signals as the basis for a student conversation rather than as direct evidence of misconduct.
FAQ
- Does Learnaway work for research papers and dissertations?
- Yes. The same behavioural signals apply regardless of document length. For very long papers, paste dominance and submission speed relative to word count are particularly informative.
- What if a student pasted from their own notes or prior work?
- That's a common and legitimate use of paste. Learnaway requires multiple corroborating signals before raising concern, and always leaves the judgement to the teacher. Paste alone is not flagged without additional context.
Want to see this on real writing?