AI FOR TEACHERS · HONEST GUIDES
The Teacher's Guide to AI Detectors: What Works and What Doesn't
If you've ever pasted a suspicious essay into an AI detector and wondered whether you can trust the answer — this guide is for you. Short version: these tools can start a conversation, but they can't end one. Here's why, and what to do instead.
First, the honest truth about AI detectors
No AI detector is accurate enough to prove a student used AI. Independent studies have repeatedly found meaningful false-positive rates — human students flagged as AI — and false negatives too. OpenAI shut down its own AI-writing classifier in 2023, citing its low accuracy. Stanford researchers found detectors disproportionately flag writing by non-native English speakers. Detection scores are clues, not evidence.
Why are they unreliable? Detectors look for statistical patterns — predictable word choices, even sentence lengths, formal structure. The problem: plenty of humans write that way, especially students who've been taught to write formal five-paragraph essays. And AI models keep changing, so detectors are always chasing a moving target.
The tools teachers actually use
Here's a fair look at the most common options. None of these is a silver bullet — but they differ in cost, workflow, and how they present results.
GPTZero
One of the best-known AI writing detectors, originally built with educators in mind. Gives an overall likelihood score plus sentence-level highlights.
Free tier for quick checks; highlights which sentences it flags so you can see its reasoning.
Like every detector, it flags human writing sometimes — especially polished, formal, or formulaic prose.
Originality.ai
A paid detector aimed at content professionals, combining AI detection with plagiarism checking.
Built for scanning lots of documents; combines two checks in one pass.
Paid per scan; designed for web content rather than student work, and still produces false positives.
Turnitin's AI indicator
Turnitin added an AI-writing indicator to its long-standing plagiarism platform, which many schools already license.
Fits into a tool your school may already use; no new workflow to learn.
Turnitin itself says the indicator should not be used as sole evidence of misconduct — and it requires a school license.
Copyleaks, Sapling, and others
A long tail of similar tools, each with its own scoring and claims of accuracy.
Useful as a second opinion if your school already subscribes to one.
Vendor accuracy claims come from the vendors themselves; independent research has found real-world accuracy is much lower.
So how should teachers check for AI? A fair five-step process
The goal isn't catching students — it's protecting learning while keeping trust. This process does both.
- 1
Treat the score as a question, not a verdict
A detector score is a prompt to look closer — never proof. Researchers, universities, and even OpenAI (which shut down its own AI classifier in 2023 over low accuracy) agree these tools are not reliable enough to accuse a student on their own.
- 2
Check the writing against the student, not the internet
Compare the piece to the student's past work. Ask them to talk you through their argument, their sources, or why they chose a particular example. A genuine author can discuss their own thinking; someone who pasted AI output usually can't.
- 3
Watch for the patterns AI actually leaves
Oddly generic examples, sources that don't exist, a sudden jump in sophistication, perfectly even tone, or citations you can't find anywhere. These are signals to start a conversation — still not proof on their own.
- 4
Have the conversation before the accusation
Lead with curiosity: 'Tell me how you put this together.' Students who used AI will often admit it in a low-stakes conversation, and you keep trust intact for everyone else.
- 5
Know your school's policy first
Many districts now have explicit guidance on AI use and AI detectors — some forbid using detector scores as disciplinary evidence at all. Check yours before you act on a flag.
The better long-term answer: design for it
The most effective move isn't a better detector — it's assignments where AI use is visible or beside the point: in-class drafts, personal reflection tied to your actual classroom, oral defenses of written work, and tasks that use AI openly with a "show your prompts" requirement. When students can use AI transparently, the cat-and-mouse game ends.
A SMARTER TEACHER INBOX.
Twice a month: one short email with practical AI moves, a copy-ready prompt, a responsible-use check, and a link to the full issue on the site. No noise, unsubscribe anytime.
Want ready-to-use classroom materials? Grab our free teacher resources → or see this month's live session →