Scanning ESL and EAL classes fairly
Nothing about the process changes for an ESL or EAL class. What should change is how much weight you put on the number by itself.
Know the research before you scan the first essay
Independent research has found that widely used AI detectors consistently misclassify writing by non-native English speakers, in some studies flagging a majority of honest essays from learners while barely flagging native-speaker writing at all. That is not a Litimus-specific problem, but it is not one to assume away either.
Set the baseline from this class, not a general one
A learner’s earlier work in your own class — not a generic “average student” expectation — is the fair comparison point. Sentence variety and vocabulary range grow over a course; a score that looks high in week two can look ordinary by week ten against the same student’s own trajectory.
Weight the conversation more than the number
For this population specifically, treat the score as a prompt to look closer, not a threshold to act on. The five-step process still applies — read the work yourself first, then compare against baseline, then talk to the student — but the gap between “number” and “decision” should be wider here than for a class where false positives are rarer.
What to tell the class up front
Say plainly, before any assignment is due, that detectors are known to score non-native writing differently, and that this is factored into how you read results. Knowing that in advance changes how a flagged score lands for a student who was already anxious about it.