We’ve released a preliminary version of our in-house AI detector. It lets you keep checking your documents when AI reports from Turnitin aren’t available, and it’s built to match Turnitin’s results as closely as possible.
It’s a working model, not a finished one. We’ll keep improving how it scores documents as we compare results and hear from you.
Why we built our own detector
Getting AI reports straight from Turnitin has become less reliable. When that access drops, you can lose the chance to check a paper right when you need to.
Our in-house model gives us another way to process your document. For now, we’ll use it whenever Turnitin AI reports are unavailable, so you can keep uploading as close to 24/7 as possible.
Results from our detector are Academi.cx predictions, not AI reports generated by Turnitin.
What to expect from this release
We’re aiming for about 85% accuracy against Turnitin. On most of the documents we’ve tested, the model is already at or near that mark. It’s still early, so results will shift as we keep training it.
Your overall score and the sentences the detector flags may not line up exactly with Turnitin. We’re using those differences to decide what to fix next.
What comes next
We’ll keep training the detector and post updates as changes go live. Each update is aimed at bringing our results closer to Turnitin’s.
If you try the new detector, tell us how it does on your documents. Feedback on the score and the flagged sentences shows us where it needs work.
