On February 12, 2026, we noticed that Turnitin had updated how it scores documents for AI. In simple terms, Turnitin became less harsh on AI writing, so documents started getting lower AI scores than before.
Our in-house detector is built to match Turnitin’s results, so for a few days our scores ran higher than Turnitin’s. Here’s what changed and how we updated our detector to catch up.
What changed
Turnitin didn’t announce the update. We spotted it by comparing results: documents that went through both systems came back with lower AI scores on Turnitin than on Academi.cx.
At first we weren’t sure the change would stick. It’s unusual for a detector to suddenly get less strict, and it could have been a short test or a bug. As more results came in, it was clear the update was here to stay.
How we updated our detector
On February 15, we released a preliminary update to our detector to account for the change, while we kept collecting examples of how Turnitin was now scoring documents.
A second update followed on February 18. In initial testing on February 19, the detector was at about 90% sentence-level accuracy against Turnitin. You can read more about those updates in our post on sentence-level AI detection updates.
What this means for you
If you checked a document in the first few days after February 12, your Academi.cx score may have come out higher than your Turnitin score. Our February 15 and 18 updates were made to bring the detector in line with Turnitin’s new scoring, and we’ll do the same whenever Turnitin changes its scoring again.
