Can I rely on Detecto for graded-pass / fail decisions?

No, and no detector should be used that way. The result is a likelihood meant to inform a human review, not to issue a grade. Use it to decide whether a conversation is warranted, and make the academic-integrity decision through your institution’s process.

What about ESL students whose writing might score higher?

Detection isn’t language-blind, and non-native phrasing can move a score. That’s exactly why the result is evidence, not a verdict: read the flagged passages, weigh the confidence band, and confirm with the student rather than acting on the number. See our accuracy and false-positives pages for the full posture.

Is the student’s essay shared with anyone?

No. Submissions are saved privately to your account and are never public unless you create a share link. Run in ephemeral mode to store nothing. We don’t use student work to train detection models.

For educators

For educators reviewing student work

A probability and a paragraph of evidence, never a unilateral verdict.

When a student’s essay reads like a model wrote it, you need more than a feeling. Detecto returns an AI-likelihood and the evidence behind it so you can open a fair, specific conversation, not deliver a unilateral verdict. Student work stays private to your account.

See sample report How calibration works

Built for the people who review other people’s content

Editors at independent magazines
University writing centers
Newsroom verification desks
PR review teams
Content QA leads at agencies

What you review

What educators typically scan with Detecto.

Essays and analytical writing
Take-home exam responses
Lab reports and methodology sections
Application essays and personal statements

What’s at risk

The failure modes Detecto is built to surface.

A submission is generated wholesale and turned in as original work
An AI draft is paraphrased just enough to sound like a student’s voice
Genuine student writing falls under suspicion with no evidence to check
Pressure to act on a hunch instead of a documented basis

Where Detecto fits

The workflow, end to end.

Five steps. None of them is automated.

01
Paste the passage, or upload the file.
02
Read the score and confidence band, a low band means inconclusive, not cleared.
03
Open the highlighted view to see which passages drove the result.
04
Talk to the student with the evidence in front of you.
05
Keep the report and your note as a record of how the case was handled.

What the report gives you

Evidence that goes in the file.

A 0–100 AI-likelihood with a High / Medium / Low confidence band.

Sentence-level signals showing where the writing reads as machine-generated.

A private, timestamped record that documents a fair, evidence-based process.

What not to do with AI detection

The same posture every Detecto reviewer reads. The score is evidence, not a verdict.

Don’t fail or sanction a student on the score alone.
Don’t read a low band as proof of innocence, or a high one as proof of guilt.
Don’t skip the conversation; the score is the start of it, not the end.
Don’t assume non-native writing is AI, confirm with the student, not the number.

All four modalities