Proctorio Blog

Introducing AuthorProof

For twenty years, academic integrity tools have all asked the same thing: "Is this original?" They compare a submission against a database, or against a model's sense of what AI-generated text looks like, and they return a probability. For most of those twenty years, that was enough.

It isn't anymore.

A student can open ChatGPT, paste in the assignment prompt, and have a polished, well-argued, fully "original" essay in under a minute. It matches nothing in any plagiarism database, because it was written from scratch. AI detectors might flag it, or might not, and either way they can only ever offer a guess. No instructor wants to bring an academic integrity case built on "the software was 68 percent confident."

The problem isn't that the tools are bad at their job. The problem is the job itself. Originality detection looks at the document. It never looks at the student.

The blind spot

Think about everything a document-based check cannot see:

• AI-generated content. Large language models produce essays that are indistinguishable from strong student work. A detector can estimate probability. It cannot prove who sat at the keyboard.
• Contract cheating. A student can buy an essay or hand the assignment to someone else. The writing is genuinely original. It just isn't theirs.
• Paraphrasing and ghostwriting. Rewriting a source, or having a friend draft the argument, sails past every similarity check. Instructors have always known this happens. They've just never had a tool that could act on it at scale.
  In each case the document looks clean. The question "Is this original?" comes back with a reassuring answer, and the real question goes unasked.

A different question

AuthorProof asks the one thing that actually matters: "Did you write this?"

Instead of analyzing the essay, it assesses the author. When a student submits written work, AuthorProof reads that specific submission and generates a short comprehension quiz drawn from it: the arguments it makes, the sources it cites, the claims at its core.

Then it asks the student about their own work.

The logic is simple, and it holds up no matter how the essay was produced. A student who genuinely wrote about the drivers of coral bleaching can tell you why they landed on rising sea temperatures over overfishing. A student who generated the essay ten minutes before the deadline cannot, because they never engaged with the ideas in the first place.

It doesn't matter whether the text came from ChatGPT, a paid writer, or a heavily paraphrased source. If you understand what you submitted, you can answer questions about it. If you don't, you can't. That is a signal document analysis can never produce.

How it works, inside Canvas

AuthorProof isn't a separate portal to log into or a new dashboard to learn. It lives inside Canvas, in the workflow instructors already use.

1. Enable it on the assignment. AuthorProof settings appear inline in the Canvas assignment editor, under Plagiarism Review. The instructor sets the question type (basic recall or deeper comprehension), the proctored environment, the completion window, and the quiz language. No context switching.
2. The student submits, and the quiz builds itself. The moment a student turns in their written work, AuthorProof generates a comprehension quiz personalized to that exact submission. It takes seconds, and it happens automatically.
3. The student verifies their work. An email arrives as soon as the quiz is ready, with the deadline, the course, the length, and a direct link to open it in Canvas. The questions target the arguments, citations, and key claims in their essay. If they wrote it, they know the answers.
4. The score lands in SpeedGrader. The authorship result shows up as a badge right on the submission. Click it to open the full session in Review Center, where you get question-level results, a confidence rating, and the authorship detail behind the number.

The instructor never leaves the grading tools they already know. The signal is just there, next to the grade, when they need it.

Built for real classrooms

A verification tool is only useful if it works for every student, not just the convenient cases.

• Instructor control. Question type, completion window, proctoring level, and language are all configurable per assignment, in the assignment editor.
• Multilingual by default. Questions generate in the language of the submission automatically. Instructors can override to English while keeping cited passages in their original language, so a Spanish-language essay is verified in Spanish.
• SpeedGrader integration. Scores appear where grading already happens, with one click through to the full breakdown.
• FERPA compliant. Submission content isn't stored beyond quiz generation, and no personally identifiable information lands in the logs. It's designed for institutional security standards from the first day, not bolted on later.

The shift worth making

The generative AI era didn't break academic integrity. It broke a single assumption: that you can tell who wrote something by studying the writing. That assumption is gone, and no amount of better detection is going to bring it back.

The way forward isn't a more confident guess about the document. It's a straightforward conversation with the author. AuthorProof makes that conversation automatic, scalable, and native to Canvas, so instructors can stop asking whether an essay looks original and start confirming that the student in front of them actually wrote it.