For faculty

How to stop students from using AI to write their essays

No setting or detector stops it outright. What works is a combination: a rule students understand, assignments where the process counts, and a record of how each essay was written.

Last reviewed October 22, 2026. Written by the EssayGuard team.

In short

You cannot block AI from a student's life, but you can make it the harder path on your assignments. State a clear rule, design essays where drafts and choices are graded, have students write where the process is recorded, and follow up with a conversation rather than a score. A visible process makes the honest path the easy one.

Why blocking alone does not work

The first instinct is technical: block the sites, buy a detector, lock the browser. Each of those helps a little and fails in a predictable way. Network filters stop at the school’s edge, and students write at home. Detectors score finished prose, so they flag some honest students and miss some generated text, and humanizer tools exist to change the score. NBC News reported on January 28, 2026 that 43 humanizer sites drew 33.9 million visits in a single month. Locked browsers fit a timed test, not a take-home essay that needs research and revision.

Source: NBC News, January 28, 2026, checked 2026-09-08.

What changes behavior is the same thing that changes it for any rule: students know what the rule is, they can see that following it is expected and checked, and the check is fair. That gives you four layers, and they work best together.

Layer one: a rule students can follow

Most AI problems in a class start with a vague policy. “Do not cheat with AI” leaves a student unsure whether a grammar checker, an outline from a chatbot, or AI dictation counts. Write the rule per assignment and say three things: what AI use is allowed, what is not, and how the work will be checked.

  • Allowed: for example, asking a chatbot to explain a concept you then write about in your own words.
  • Not allowed: for example, pasting or retyping any text a tool wrote, including AI dictation tools.
  • How it is checked: for example, “this essay is written in a recorded editor, and I read the record when I have a question.”

Our classroom AI policy template has wording you can paste, and the syllabus language page has a college version.

Layer two: assignments where the process counts

An essay graded only on its final form invites a shortcut to the final form. When part of the grade comes from the work along the way, the shortcut buys less. A few patterns that hold up:

  • Grade the thesis, an outline and a first draft as separate steps, even at low weight.
  • Tie the prompt to something only your class has: a discussion, a local source, the student’s own earlier essay.
  • Ask for a short reflection on one choice the student made and one thing they cut.
  • Hold a five minute conference on the draft. A student who wrote it can talk about it.

The AI-resistant writing assignments guide goes deeper on each of these and what they cost in grading time.

Layer three: a record of how the essay was written

This is the layer most teachers are missing. When students write in an editor that records the session, you can see how each essay came together: what was typed, what was pasted, where the student paused, what they deleted and rebuilt. Generated text usually arrives in a paste, or it is retyped from another window in a steady rhythm with few changes. Either way it looks like copying, and it looks different from the stop and start of real drafting.

A Google Docs version history is a weaker version of this. It shows snapshots of the document, it does not separate typed text from pasted text, and auto-typer tools can replay pasted text into a Doc so the history looks written. See can students fake a writing history for how that works.

The record also works before anything is written. Students are told at the start that the session is recorded, so the rule and the check arrive together, and a student weighing a shortcut knows the process will be visible.

Layer four: a fair follow-up

A rule that is never checked stops working, and a check that feels like an ambush makes students defensive. When something looks off, start with a question about the work, not an accusation. Ask the student to walk you through the argument and one choice they made. The guide to what to do when you suspect AI has the full sequence and a first message you can send.

How EssayGuard fits

EssayGuard is the third layer. You create an assignment in about three minutes, set the rules for that piece of work (whether quotations and citations may be pasted, whether device dictation is allowed), and share one link. Students sign in and write in the browser, with nothing to install. Every keystroke, revision, pause and paste is recorded with server-side timestamps, along with when the student left the page and came back.

Each essay comes back marked Approved, Needs Review or Flagged, with a short summary of what the recording shows, so you know which replays to open first. There is no AI percentage, and any decision about a student is yours. It is free for the rest of this semester, then $99 a year. Start with one assignment.

What none of this can do

A record shows how text entered the editor. It cannot show whether someone else supplied the ideas, and a patient student can retype generated text slowly. No layer on this page makes AI use impossible. Together they make the honest path the easy one, give you evidence worth discussing, and keep the decision with a person who knows the student.

Questions people ask

Can I block ChatGPT on student devices?

A school can filter websites on its own network and managed devices, but students write at home, on phones, and on other networks. Blocking reduces casual use in class. It does not stop a determined student, so it works best as one layer alongside a clear policy and a record of the writing.

Does an AI detector stop students from using AI?

Not reliably. A detector scores the finished text, so it can be wrong in both directions, and students know that humanizer tools exist to change the score. A detector percentage is at most a reason to look more closely. It should never be the finding on its own.

Should I ban AI completely?

That is your call and your school's. Many teachers allow AI for brainstorming or feedback and forbid it for drafting. Whatever you choose, write it down per assignment, because students mostly get into trouble where the rule was vague.

Do in-class essays solve the problem?

They remove most AI use for that piece of writing, at the cost of time, research, revision and students who need accommodations. Many teachers keep one timed piece and record the take-home essays instead.

Try it with one assignment

The rest of this semester costs nothing. Setup takes about three minutes.

Every course you teach, unlimited students. Students install nothing.