AI Detector for Teachers

A review workflow, not a verdict

Detection should support review, not replace human judgment. The checker is built to give you something to read and discuss, so the page walks through four steps.

1. Check student writing. Paste the text or upload a TXT, DOCX or PDF file. Include the full piece where you can; very short samples give weak results.
2. Understand the result. Read the verdict, the confidence level and the share of the text flagged as AI-like before looking at any highlights.
3. Review flagged passages. Go through the highlighted sentences in context. Each passage is labeled AI-Generated, AI-Assisted or Human Written, with High, Medium or Low confidence.
4. Have an informed conversation. Bring the passages, the student’s earlier work and any drafts to a calm, open discussion about how the piece was written.

Interpreting what you see

“Flagged as AI-like” is the share of the text made up of passages the detector labeled AI-like. It is not a probability that the student used AI, and it is not a grade for honesty. A figure of 30 percent means roughly three tenths of the passages were labeled AI-like, which is a prompt to read those passages, not a conclusion.

• Mostly human-like usually needs no action, though it does not prove a piece was unassisted.
• Mixed signals is where reading the highlights matters most. Look at whether flagged sentences cluster together and how confident each label is.
• Likely AI-assisted deserves a closer look at process: drafts, notes, sources and the student’s own account.
• Not enough text means the sample is under 50 words. Ask for more writing before drawing anything from it.

The detector tells you which passages were labeled, not why. We do not guess at reasons, and you should not either.

Before you raise a concern

Never act on a detector result alone, and never use it as the sole basis for a penalty. False flags happen, and the cost of one falls on a student. Before you speak to anyone, gather other context:

• Earlier writing by the same student, to compare voice and level.
• Drafts, outlines, notes or version history that show how the piece developed.
• The assignment’s own rules on AI tools, since expectations differ between courses and schools.
• Whether the student writes in a second language or uses assistive tools.

In the conversation, ask open questions: how did you start, what sources did you use, can you explain this paragraph in your own words? Your school’s academic integrity process, not this page, decides what happens next. For more on how results are produced, see the methodology, and for the limits we can and cannot claim, the accuracy page.

Handling student work responsibly

We keep the privacy facts plain. Our server does not store the text you submit or any uploaded file, and our logs record only a request ID, the action, duration, provider, outcome and word count, never the text itself. Analytics are anonymous events such as a word-count range and a verdict category.

The text is, though, sent to Pangram Labs so it can be analyzed. We do not know Pangram’s retention practices and do not claim anything about them; its own terms apply. If your school restricts where student work can be sent, follow that policy, and consider removing names before you check. Full details are on the privacy page.

Related