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

    1. 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.

    2. 02

      Use a suitable sample

      Avoid drawing conclusions from very short extracts, reference lists, quotations, formulaic templates or text outside the supported language.

    3. 03

      Read the detailed result

      Look at sentence-level signals and source matches rather than relying on a single percentage or label.

    4. 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.

    5. 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.

    6. 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.

    Examples of evidence in an academic-integrity review
    EvidenceWhat it can showImportant caution
    Drafts and version historyHow an argument and wording developed over timeNot every genuine writer keeps complete records
    Notes and source trailHow evidence was selected and understoodNotes can be incomplete or produced in different tools
    Student explanationWhether the writer can explain choices, claims and revisionsUse neutral, accessible questions and a fair procedure
    AI detection resultWhich passages resemble patterns the detector associates with AI textFalse positives and false negatives are possible
    Source-matching resultWhere wording overlaps with sources found by the providerA 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.