New Tool Shows How Students Actually Use AI in Their Writing
Ninety percent of college students use AI in their coursework, with nearly half incorporating it during the drafting stage. Yet professors struggle to understand what that collaboration actually looks like. A new open-source tool called DraftMarks, developed by researchers at Georgia Tech and Stanford, makes the writing process visible by marking where AI was involved, how heavily it was used, and what a student chose to keep or discard.
DraftMarks functions as a reading layer that sits on top of a document. Instead of flagging AI's presence with a percentage score, it uses visual cues that mimic physical artifacts of traditional writing - eraser crumbs, tape, smudges, and smudged text - to show different types of AI interaction.
What the marks mean
Eraser marks indicate heavily revised passages. Masking tape highlights text initially generated by AI. Smudges signal where AI changed the strength of an argument rather than just the content. Glue residue shows where AI-generated text was later removed. Ghost text appears when a writer prompted AI but rejected the output. Different fonts distinguish human-written passages from AI-generated ones.
Together, these marks tell a story about the writer's decision-making process, not just whether AI was used.
How researchers designed it
The team started by observing 21 instructors reviewing student writing. They watched what cues educators already looked for when assessing learning, revision, and originality. Those insights shaped DraftMarks' visual language.
"These marks emulate the writing process in ways we're already familiar with," said Adam Coscia, a computing Ph.D. student on the project. "They help students and teachers see the effort behind the writing, and whether students actually met the learning objective."
The tool tracks a document's draft history and classifies different types of edits and AI interactions as they happen, allowing the visual cues to appear almost in real time.
What readers found
A follow-up study with 70 participants - including students, teachers, journalists, and general readers - revealed different priorities. Instructors focused on how ideas developed and where students exercised judgment. General readers used the marks differently, assessing trust and authorial intent rather than the learning process.
For writers, seeing the marks shifted perspective. One researcher said the tool changed how he thought about his own writing. "I was surprised by how much I cared about authorial intent once I could actually see how AI affected my tone. It made me realize small AI choices can subtly reshape what I'm trying to say."
A shift away from detection
Unlike AI detectors that produce a percentage score, DraftMarks prompts reflection. It doesn't answer whether AI was used - it asks writers and readers to think about why certain choices were made.
As AI becomes embedded in everyday writing, tools designed around detection fall short. The research team hopes DraftMarks will shift conversations toward transparency, giving educators and writers a clearer window into how learning happens when humans and AI write together.
The work was funded by the National Science Foundation's AI Research Institutes program and the Institute of Education Sciences at the U.S. Department of Education. The team presented DraftMarks at the Association for Computing Machinery's Conference on Human Factors in Computing Systems in Barcelona in April 2026.
Your membership also unlocks:
