Classroom-grade reports
PDF export per essay, cryptographically signed, suitable for disciplinary panels. Every claim is backed by specific sentence evidence.
For Teachers · Integrity without burnout
Evidence you can take to a conversation.
Stop defending a single number. Our report gives you a sentence-level heatmap, a model-family fingerprint, and a hybrid-draft ratio, the three things you actually need when you sit down with a student.
Scan a sample essay → Request classroom pricing
- Academic AUC
- 0.94
- Faculty using us
- 8,200+
- LMS integrations
- 4
Why faculty ask for evidence, not verdicts.
We've talked to hundreds of faculty about AI detectors. The complaint is nearly universal: the tools give you a single percentage score, the student disputes it, and you're stuck in a meeting with no evidence beyond "the computer said 87%." That's not enough to hold up in an integrity hearing, and, more importantly, it's not enough for the conversation that should happen first.
We built this around a single idea: evidence beats verdicts. Every scan produces sentence-level highlights, a model-family fingerprint (GPT-4o? Claude 4? Gemini 2.5?), and a hybrid-draft ratio that separates "AI-assisted editing" from "AI-drafted submission." That's what an integrity conversation needs.
We also built it to not replace your judgment. Our reports surface evidence and ask you to decide. We don't auto-fail, we don't auto-report, and we don't generate a canned accusation. The tool is yours.
What you get as faculty.
LMS integrations
Canvas, Blackboard, Moodle, D2L. Essays never leave your institution's tenant. Bulk-scan an assignment in one click.
Hybrid-draft scoring
Differentiate 'student used GPT-4o for one paragraph and edited the rest' from 'student pasted a full model output.' Different conversations, different responses.
Rubric integration
Attach our AI-likelihood score as an optional rubric dimension, weighted however you want, or excluded from grading entirely.
False-positive protection
We flag low-confidence scores loudly and recommend against formal action below 0.80 probability. You shouldn't be accusing anyone on a coin flip.
Student-facing version
An optional 'for student' view that shows the same heatmap in plain language. Great for a one-on-one conversation that isn't confrontational.
A classroom workflow we recommend.
1. Announce it up front.
Before the first assignment, tell students you use an AI detector and show them the tool. Transparency cuts the 'gotcha' framing and dramatically reduces cheating rates.
2. Scan in bulk via LMS.
When the assignment closes, run the full batch. Our LMS integrations return a ranked list, focus your attention where the signal is strongest.
3. Review the evidence before you act.
For high-probability flags, open the report, read the heatmap, check the model fingerprint. Look for hybrid-draft scores, those change the conversation.
4. Start with a conversation, not an accusation.
Share the student-facing heatmap. Ask them to walk you through how they wrote the flagged sections. Many cases resolve here without escalation.
5. Escalate only with evidence.
If it needs to go to the integrity office, export the signed PDF. It contains everything the panel needs, methodology, confidence band, sentence evidence.
What faculty get
Built for the people who actually grade essays.
1.8s
Per-essay scan
Bulk class sets in under three minutes.
CSV
Round-trip export
Drops cleanly into Canvas, Blackboard, Moodle, D2L.
PDF
Hearing-ready
Cryptographically signed, sentence-level evidence.
<3%
ESL FPR
Audited fairness on 12,000 non-native essays.
Frequently asked questions
How do you avoid false positives ruining a student's record?
Three ways. (1) We surface confidence bands, not hard verdicts, below 0.80 probability we explicitly recommend not acting. (2) We flag non-native-English patterns so a teacher can consider that before escalating. (3) Our report emphasizes that it's evidence for a conversation, not proof for a verdict. We built this because the biggest complaint from faculty was false positives on ESL students.
What if the student disputes the score?
The report's sentence-level heatmap is the starting point. Ask them to walk you through the flagged sentences. If they can explain their reasoning and revise in your presence, that's usually enough. If they can't, that's also information. Either way, the conversation is more productive than arguing about a number.
Do you integrate with Turnitin or Gradescope?
We have plugins for Canvas, Blackboard, Moodle, and D2L that run alongside Turnitin. We don't directly replace Turnitin, their plagiarism corpus is different from ours. Many institutions run both: Turnitin for plagiarism-source-matching, us for AI detection.
Is there a per-teacher subscription, or is it institution-wide?
Both. Individual faculty can subscribe ($9/mo with reasonable scan limits). Institutions get per-seat pricing, bulk scan, LMS plugins, SSO, and a DPA. See the institutions page for enterprise details.
What does the PDF report contain for an integrity hearing?
Essay text, sentence-by-sentence AI probability, perplexity and burstiness scores, model-family fingerprint, hybrid-draft ratio, methodology reference, timestamp, and a cryptographic signature. Every data point is explained in a glossary appendix so a non-technical panel can read it.