How Our Flags Work
TrueMark automatically flags noteworthy content in student writing through two distinct methods:
Pasted content (highlighted) - When students copy and paste text.
Transcribed content (underlined) - When we identify suspicious typing patterns based on irregular insertion/deletion and typing speed across 6 or more sentences. This can uncover work that may have been transcribed from an external source.
Understanding AI Flags
We'll mark noteworthy content as AI-generated (in red) if two different detection models indicate >90% confidence, and in yellow if not.
As of March 2025, these AI detection models have a 2% chance of false positives.
Traditional web plagiarism flags whether the student's work may have been pulled from an external online source.
How to Review Flagged Content
In the summary report, hover over the flag(s) to see what content triggered detection.
When reviewing the essay, click on highlighted or underlined text to view exactly what content was flagged.
Use the student playback feature and click on the red dots on the timeline to see the exact moments AI usage was detected.
The AI usage percentage represents the portion of final text flagged as AI-generated.
This allows you to easily identify when students are using text directly from another source, whether pasted or manually typed.
Trying to navigate a specific integrity report? Walk through this step-by-step guide
Our AI-detection philosophy
AI detection methods are not perfect, and their accuracy is getting worse over time.
AI-detection models quickly become out-of-date as new generative AI models get released.
Students frequently use AI-generated content as a starting point (e.g, an outline or a couple of paragraphs) and modify it to obscure its AI origin.
AI-detection models can have false positives, which can make them unreliable.
Our goal is to have low-volume, higher-confidence positives and offer transparency into the student writing process. To do so, we primarily check suspiciously typed content against multiple different AI-detection tools. This allows us to:
Capture a student’s entire edit history, not just their final outcome
Improve the reliability of AI-flagged results
Provide anomaly detection for student work that AI detection tools may have overlooked, enabling educators to use their own expertise to assess academic integrity.
Need additional assistance? Contact our support team for further guidance.
FAQs
Q: I noticed that flags are tagged with labels like "internal", "external", "copilot", and "unknown." What do they mean?
When students write within the TrueMark editor, we capture additional data on where certain flags come from.
Internal: content was copied and pasted from within their own document. Most frequently, this is due to students moving paragraphs around during the editing process.
External: content was pasted in from outside the document. As a reminder, you can disable external pastes on your assignment if you don't want students to be able to paste external content into their document.
Copilot: content was pasted into the document directly from TrueMark's copilot. You can verify this by looking at the student's chat history -- the pasted content should also live there.
Extension: more than 10 words of content was inserted by a browser extension (e.g., Grammarly, translation tools, AI writing assistants, etc.). This suggests the student is using a browser tool to generate or modify text. This could be a grammar checker (benign) or an AI writing tool (potentially concerning depending on your policies).
Unknown: we couldn't determine where this content was pasted from. This can happen due to one of a few reasons:
The document was created before October 2025
The student used "paste and match style" instead of "paste"
The document is a Google Doc. Detailed pasted information is only available within the TrueMark editor.