Prove the code is authentically theirs.
A code authenticity check answers one question: did this student write this code? Codequiry puts three kinds of evidence in one report: how the code was produced, where else it exists, and whether it reads as AI-written.
Runs on any upload. Add the Window editor for a full writing record.
Copied code has a source. AI-written code does not.
A plagiarism checker looks for the place code came from. Code a model wrote for one student exists nowhere else, so there is nothing to match. Authenticity has to be shown another way: by how the file was built, and by what the code itself looks like.
How it was written
In the Window editor every paste is recorded with its size and time, and minutes spent outside the editor are timed. A block that appears in one paste right after a student comes back is flagged.
Where else it exists
Each file is compared with every classmate and with GitHub, Stack Overflow and the web. Every match links to its source.
Who likely wrote it
Every file is scored for AI generation, with the regions that drove the score highlighted and an estimate of the model family.
Evidence you can put in front of a student.
Each signal is shown with what produced it, so a conversation starts from facts on the screen instead of a number nobody can explain.
A writing timeline
Window records the session as the student works: typing, each paste and its size, time away from the editor, and a screen capture when a paste lands.
- Share of the file typed by hand
- Pastes that land right after returning
- Time off-window, minute by minute
AI authorship, per file
A probability for each file, the windows of code that read as generated, and a tentative model family such as GPT, Claude or Gemini.
- Flagged regions, not just a score
- Model family estimate
- Included in every check
Peer comparison
Every submission against every other one in the course, robust to renamed variables and reordered functions.
Web and GitHub sources
Matches against public repositories, Stack Overflow and tutorial sites, each linked to the page it came from.
Side-by-side diffs
Matched lines highlighted in both files, ready for an academic integrity meeting.
From submission to a clear answer in four steps.
Students write the code
In the Window editor for a full writing record, or anywhere else and upload as usual.
Run one check
Peer, web and AI detection run together on the same upload.
Read the evidence together
Process, sources and AI signals sit side by side for each file.
Talk to the student
Open the conversation with the record on screen, not with an accusation.
One number is not proof. Agreeing signals are.
An AI score is a probability, and no detector is right every time. Authenticity is decided when independent signals point the same way: a file that reads as AI-written, arrived in one paste, and landed after minutes away from the editor.
Code authenticity, answered.
What is a code authenticity check?
It is a review of whether a person actually wrote the code they submitted. Codequiry combines three kinds of evidence: how the code was produced (from the Window editor), whether it matches classmates or public sources, and whether it reads as AI-written.
Can you prove a student wrote their own code?
No single test proves authorship. What a report can show is whether independent signals agree: a writing record with steady typing and no large pastes, no matches to other sources, and a low AI score point to authentic work. The opposite pattern is a reason to talk to the student.
Does it work without the Window editor?
Yes. Peer comparison, web and GitHub checks and AI detection run on any upload. Window adds the writing record, which is the strongest evidence for code that has no source anywhere else.
How accurate is the AI detection?
In held-out testing, model v3e caught 90% of AI-written code and falsely flagged 1.3% of human-written code. Treat the score as evidence to weigh with the other signals, not as a verdict on its own.
Do students know what Window records?
Yes. Window is a student editor that states what it records: pastes with their size and time, typing, and time away from the editor. It is a condition of the assignment, not something hidden.
Know whose code you are grading.
Run your next assignment through a check that shows how the code was written, where it came from and whether a model wrote it.
More on AI-written code
Product names such as ChatGPT, Claude, GitHub Copilot, Cursor and Gemini are trademarks of their respective owners. Codequiry is an independent service and is not affiliated with, endorsed by, or sponsored by any of them.