If your programming course runs on Gradescope, it is natural to wonder whether it can tell when code came from ChatGPT or another assistant. The short answer is that Gradescope's documentation does not describe any AI code detection. What it does describe is a similarity tool that compares students with each other, an autograder that runs your instructor's tests, and a full record of your submissions. Those are worth understanding, because they shape what your instructor can see. Everything below about Gradescope comes from its official help guides, which we read in October 2026; check the linked pages for the current wording.
What Code Similarity actually does
Gradescope's guide to generating a Code Similarity report says the feature is available for courses with an Institutional license and only for Programming Assignments. It describes Code Similarity as a tool to help determine how similar students' code is to each other, and it says plainly that it does not automatically detect plagiarism. Instead, it shows how similar two programs are within a single assignment.
To run a report, an instructor picks a programming language and can upload starter or template files so that matches against that shared code are ignored. The report lists each student's files alongside the other student whose submission was most similar. By default, results are sorted by the length of code that was similar across the pair, a number the guide says is related to the count of tokens (keywords, identifiers, operators, literals and so on) in the matching passages. Clicking a file opens a side-by-side comparison with numbered buttons to jump between matching blocks.
The guide lists the languages Code Similarity can review: C, C++, C#, F#, Go, Java, JavaScript, Matlab, MIPS, ML, Python and R. The comparison is between students in the same assignment, not against the web and not against a model.
Why similarity is not AI detection
A language model generates a new answer each time it is asked, so two people asking for the same function can get different code. A similarity report answers one question: which submissions in this assignment look like each other? It does not answer "was this written by a person or by a model?". That is the same limit as MOSS and JPlag; see does MOSS detect AI code for how those tools describe themselves.
Similarity can still matter for AI use in one case: if several students submit closely matching generated code for the same prompt, the overlap shows up like any other match. The report still cannot say why the code is similar. Someone has to look and decide.
What the autograder sees
Gradescope's guide to grading a programming assignment explains that instructors provide a setup script and an autograder script, and Gradescope runs that autograder on each submission and returns the results. The autograder checks whatever your instructor wrote it to check, usually whether your program passes test cases. Gradescope's documentation does not describe the autograder judging who or what wrote the code.
Instructors can also grade code manually with a rubric and in-line comments, combine manual and automatic grading, rerun the autograder on all submissions, and export every student's code in bulk. So even when the score comes from tests, a person may still read your code. The guide to creating a programming assignment notes that students can submit files directly or through GitHub and Bitbucket.
What your instructor can see
The student guide to submitting a code assignment says you can resubmit as many times as you like before the deadline, unless your instructor limits it. It also says your instructor sees the score for your active submission by default and can view and activate any of your other submissions from Submission History. In practice that means your instructor can see:
- Your code itself, for every submission, not just the last one.
- Your autograder results, including which tests failed along the way.
- The pattern of your submissions, since each one is recorded with its result.
- Similarity matches with classmates, if they run a Code Similarity report.
None of these is an AI flag. But they are the kind of context an instructor uses when something looks unusual, for example when a submission is very different in style from your earlier work.
What about Gradescope's AI features?
Gradescope does use AI, but for grading, not detection. Its guide to AI-assisted grading and answer groups describes AI that suggests groups of similar answers on fixed-template PDF assignments so an instructor can grade them together. The supported question types are manually grouped, multiple choice, math fill-in-the-blank and text fill-in-the-blank. That guide does not describe checking whether a submission was written by AI.
What this means for you
- Know your course's AI policy. Gradescope is just the submission platform. Whether you may use an assistant, and how to disclose it, is set by your instructor and school.
- Keep your process visible. Commit often if you submit through GitHub, keep drafts, and do not be afraid of early failed submissions. They show how your solution grew.
- Understand everything you hand in. Many courses ask students to explain or modify their code. Being able to do that is the strongest evidence of your own work.
- Do not panic over a similarity match. It is a starting point for a conversation, and Gradescope itself says it is not a plagiarism verdict.
For the wider picture of how instructors notice AI-written code without any special tool, read can professors tell if you used ChatGPT for code.
Checking how your code reads
You cannot run a class-wide similarity report on your own work, but you can see whether your code has the style patterns people associate with AI output. Paste it into the detector at the top of this page, read the reasons it gives, and compare them with how you really write. Treat it as a way to understand what a grader might ask about, not as a verdict, and be ready to explain any part of your solution in your own words.
Read more
- My code was flagged as AI but I wrote it — steps to take if a tool or grader flags your work.
- Does Turnitin detect AI code? — what Turnitin's documentation says about code.
Frequently asked questions
Does Gradescope have an AI code detector?
Not according to its own guides. Gradescope's help pages describe Code Similarity, which compares students' programs with each other, and an autograder that runs the instructor's tests. Neither is described as detecting whether code was written by AI. Gradescope's AI features are for grouping and grading answers, not for spotting AI-written work.
Can Code Similarity catch ChatGPT code?
Only indirectly. Code Similarity shows how similar two submissions in the same assignment are. If two students hand in closely matching AI-generated code, the overlap can show up. A single AI-written solution that does not match anyone else's will not stand out on similarity alone, and the tool cannot say where any code came from.
Can my professor see all my Gradescope submissions?
Yes. Gradescope's student guide says your instructor sees the score for your active submission by default and can also view and activate any of your other submissions from Submission History. Every resubmission you make is part of that history.
Is a high similarity score proof of cheating?
No. Gradescope's guide says Code Similarity does not automatically detect plagiarism; it shows how similar two programs are. Short tasks, starter code and standard algorithms often produce similar code, so a person has to look at the matches and decide what they mean.
My Gradescope submission was flagged. What should I do?
Stay calm, gather your drafts and commit history, and ask what exactly was flagged. If it is a similarity match, explain how you built your solution. If someone suspects AI use, offer to walk through your code and answer questions about it. Check your course's AI policy so you know what was allowed.
Try the detector
Have a snippet you are unsure about? Paste it into the free AI Code Detector to get a verdict, an AI probability and the reasons behind it. No sign-up, and your code is not stored.