AI grading tool flagged half my essays as AI written and it was a mess

I teach 8th grade English and last week our district rolled out a new AI detection tool for grading. I uploaded 30 student essays and it flagged 15 of them as AI generated, including one I watched a kid write in class. Turns out the tool was triggering on common transitional phrases like "in conclusion" and "firstly." I spent a whole afternoon manually reviewing each flagged paper and had to email parents explaining the false positives. The next day I found a setting to adjust the sensitivity, and it dropped the false positives down to 2. Has anyone else dealt with an AI detector that can't tell the difference between a middle schooler and a robot?

2 comments

2 Comments

abby_robinson581mo ago

OH COME ON, @jessica_miller, you're totally missing the point. The district rolled this thing out expecting teachers to just USE it, not spend hours fiddling with settings nobody told us about. I teach writing too and I've seen this exact mess play out - these tools punish kids for using basic structure they learned in class. You can't blame a teacher for trusting a tool the district pushed on them without warning. The real problem is these companies selling detectors that can't tell a 13 year old's "firstly" from a GPT output, and we're the ones cleaning up the damage. This isn't about me being careless, it's about software making promises it can't keep and wasting OUR time.

jessica_miller1mo ago

Honestly I kinda disagree here. If your tool was on default settings and you didn't test it first that's more on you than the software.