A Five-Point Plagiarism Checklist for Christian Schools

One of the most common questions I get from school leaders is: "How do we catch students who are using AI to cheat?" The honest answer: it's harder than you think, and the tools we have aren't as reliable as we'd like. AI plagiarism detectors are about 98% accurate individually, which sounds good until you realize that in a school of 500 students, that 2% error rate means you could be falsely flagging 10 students per assignment cycle. And here's the equity dimension that should trouble every Christian school: the students most likely to be falsely flagged are non-native English speakers. Their writing patterns are more likely to trigger AI detectors even when they've done their own work. In a school community committed to justice for the vulnerable, that's not an acceptable risk. At the Stony Brook School, we developed a five-point checklist with a three-of-five threshold. The goal: catch actual plagiarism while protecting innocent students from false accusations. Because the damage done by falsely accusing a student is far greater than absorbing being duped a few times. ## The Five Checkpoints ### 1. Two Independent Plagiarism Checkers Flag the Paper A single detector flagging a paper is one data point. Running it through two different detectors (Turnitin and GPTZero, for example) increases confidence. But even two flags together are not sufficient grounds for discipline. This is one checkbox, not a conviction. ### 2. Suspicious Version History We require students to write typed, at-home essays in Google Docs and to share the file with us immediately with edit access. This gives us full version history. If five polished paragraphs appear at 9:52 p.m. with no preceding drafts, that's suspicious. A typical student writing process shows false starts, deletions, reorganization, gradual development. A copy-paste job doesn't. ### 3. The Student Cannot Orally Defend the Work This is one of the most reliable indicators and the hardest to fake. If a student can't explain their argument, defend their word choices, or discuss the topic without looking at their paper, something is off. We've been doing much more oral assessment and oral follow-up to written work as a result of AI. ### 4. The Paper Doesn't Compare Favorably to Prior Writing Samples This is why schools need portfolio systems. Year-over-year writing samples allow you to track progress and spot anomalies. If a tenth grader suddenly produces prose that reads like a graduate student's, and that quality doesn't match anything in their portfolio, you have a data point. AI can actually help here: you can use it to analyze the portfolio and surface inconsistencies. ### 5. Multiple AI-Characteristic Patterns AI-generated text has telltale signs: certain word choices, syntactic patterns, a tendency toward even paragraph length, particular hedging phrases. Sophisticated students can mask these, but most don't bother. They copy, paste, and submit. If you learn to recognize the signatures of the major models, you'll catch the lazy shortcuts. ## The Three-of-Five Threshold If a paper meets three of these five criteria, any combination, we pass it to the deans for investigation. If it only triggers two, we make note of it and move on. This means that yes, some students will get away with it for a while. That's a feature, not a bug. These students are almost always repeat offenders. They'll get bolder, sloppier, and eventually they'll trip the three-of-five threshold. The alternative (treating a single plagiarism detector flag as proof) will break trust with students, create equity problems for vulnerable populations, and make teachers into accusers rather than advocates. That's not the posture a Christian school should take. ## A Pastoral Posture The checklist is a tool, not a weapon. Its purpose is to protect the integrity of the learning process while treating students with the dignity they deserve as image-bearers. False accusations damage trust in ways that are difficult to repair. A student who is wrongly accused of cheating may never fully trust that teacher or that institution again. When the three-of-five threshold is met, it triggers an investigation, not a conviction. We talk to the student. We seek to understand. And if plagiarism is confirmed, the disciplinary process is pastoral: accountability paired with a path back. I'd encourage any Christian school to adapt this framework to their own context. The specific checkpoints can be adjusted. What matters is the principle: multiple independent data points, an equity-conscious threshold, and a posture that would rather absorb being deceived than unjustly accuse the innocent. That's wisdom, not naivete.