For teachers and educators
Review AI-assisted coursework with evidence and care
Use automated signals to decide where to look—not to replace your policy, subject expertise or a fair conversation with the student.
Before detection
The strongest review process starts before submission
A detector cannot repair an unclear assessment rule. Set expectations for permitted assistance, disclosure and evidence of process first.
Name permitted uses
Explain whether brainstorming, translation, proofreading, coding assistance or generated drafting is allowed for this task.
Define disclosure
Tell students what they should record or acknowledge, and provide a simple format for doing so.
Design for process visibility
Use staged drafts, source notes, brief reflections or short discussions where they serve the learning outcome.
Plan a consistent response
Decide who reviews a concern, what evidence counts, how a student responds and how an appeal works.
A proportionate workflow
Move from signal to context in six steps
- 01
Check the applicable rule
Identify what the assessment allowed. The same use of AI can be permitted in one task and prohibited in another.
- 02
Use a suitable sample
Avoid drawing conclusions from very short extracts, reference lists, quotations, formulaic templates or text outside the supported language.
- 03
Read the detailed result
Look at sentence-level signals and source matches rather than relying on a single percentage or label.
- 04
Seek disconfirming evidence
Check drafts, notes, version history, earlier work and sources for explanations that challenge the initial suspicion as well as those that support it.
- 05
Invite an explanation
Ask neutral questions about the argument, evidence, revisions and tools used. Do not present a detector score as an established fact.
- 06
Record the human decision
Document the policy, reviewed evidence, student response, limitations and reasoned outcome separately from the automated report.
Evidence hierarchy
Give more weight to evidence that shows the writing process
No single item is universally decisive. The value of each depends on the task, context and institutional procedure.
| Evidence | What it can show | Important caution |
|---|---|---|
| Drafts and version history | How an argument and wording developed over time | Not every genuine writer keeps complete records |
| Notes and source trail | How evidence was selected and understood | Notes can be incomplete or produced in different tools |
| Student explanation | Whether the writer can explain choices, claims and revisions | Use neutral, accessible questions and a fair procedure |
| AI detection result | Which passages resemble patterns the detector associates with AI text | False positives and false negatives are possible |
| Source-matching result | Where wording overlaps with sources found by the provider | A match is not automatically plagiarism |
Student data
Check whether submission is appropriate before uploading work
Submitted text may contain personal data, unpublished ideas or confidential material.
Genutext processes submitted text through an external analysis provider and does not retain the text in scan history. Score summaries and operational metadata are retained, while detailed paid results remain temporarily in the browser unless an email report is requested.
Follow your organisation's approved-tool and data-protection rules. Remove unnecessary identifiers where practical, and review the privacy policy before submitting student or research material.
Pay as you go
Use a scan when it adds useful context
Start with a free document-level preview or choose a paid credit for longer text, sentence-level results and optional source matching.