Enterprise can scan submissions for AI-generated content and plagiarism using a pluggable detection provider. Results are private to graders (students never see plagiarism scores) and surface in three places: an inline panel in the grading interface, a detail modal for each submission, and a cohort-wide Plagiarism Report under Reports.

Detection runs asynchronously after submission so students don’t wait. Graders see results when they’re ready.

Choosing a Detection Provider

Three providers ship in Assignment Enterprise 2.1. You pick one as the active provider for your site; switching later is supported and your API keys for each provider persist independently.

• Winston AI (default on fresh installs) — both AI content detection and plagiarism, via a public self-serve API. 2,500 free credits on signup. The only provider in the registry that delivers both signals over a fully self-serve API.
• Originality.ai — both AI content detection and plagiarism. API access requires Originality.ai’s Enterprise subscription, so this is the right choice if your organization already has it.
• GPTZero — AI content detection only. GPTZero’s plagiarism checker exists as a dashboard feature but is not exposed via API, so this provider supports AI detection but not plagiarism. The Plagiarism Settings tab shows a callout when GPTZero is active to make this limitation visible.

On a fresh install, Winston AI is selected by default with full AI + plagiarism coverage out of the box. Existing customers keep their existing provider through the upgrade.

Configuring the Provider

Open Assignments > Settings and select the Plagiarism tab.

Setting Up Your API Key

1. Choose your provider from the Provider list.
2. Click the Sign up for {provider} link if you don’t already have an account, or click Get your API key if you do.
3. Paste your API key into the field below the dropdown.
4. Click Save Settings. The key is masked in the UI after saving.
5. Click Test Connection to verify that the key is valid and the provider is reachable.

Thresholds

Two thresholds control when a submission is flagged in the results UI:

• AI threshold — the AI-likelihood score above which a submission is flagged. Always visible; applies to all three providers.
• Originality threshold — the plagiarism score below which a submission is flagged. Hidden when GPTZero is the active provider, since GPTZero doesn’t return an originality score.

Submissions above the AI threshold or below the originality threshold appear in the flagged-submissions tables in the Plagiarism Report and in the per-submission grading panel.

Retention

Full vendor response payloads are kept for a configurable retention window (default 180 days). After that, the per-submission summary remains forever, but the full response payload is purged. A daily WP-Cron job handles the purging.

The summary row contains the AI-likelihood score, originality score, source-match count, flagged status, and the provider that produced it. The full response is only needed for forensic deep-dives in the per-submission detail modal.

Running a Plagiarism Check

Auto-Check on Submission

Each assignment has its own auto-check toggle:

1. Open the assignment editor.
2. Find the Plagiarism Detection section.
3. Toggle Automatically check submissions for this assignment on.
4. Save the assignment.

With auto-check on, every new submission to this assignment queues a plagiarism check immediately. There’s no global “auto-check everything” setting; every assignment is opt-in. This keeps API costs predictable and avoids surprise scans on assignments where plagiarism detection isn’t relevant.

Reading Results in the Grading Interface

When a check completes, the grading interface shows an inline panel next to the document viewer with:

• AI likelihood — the canonical 0–100% score that every provider populates. Higher means more likely AI-generated.
• Originality score — for Winston AI and Originality.ai submissions only. Higher means more original content.
• Sources matched — count of matched sources from the plagiarism check (Winston AI and Originality.ai).
• Confidence category — Human, Mixed, or AI (GPTZero only).

Click View full report to open the per-submission detail modal, which renders the complete vendor response payload (within the retention window). Click Insert summary into feedback to append a brief plagiarism summary to the feedback field, ready for editing before you save the grade.

The Plagiarism Report

Navigate to Assignments > Reports and click Plagiarism Report to see cohort-wide results. The report includes:

• Stat cards across the top (total checks, flagged, average AI likelihood, etc.)
• AI likelihood distribution chart — always rendered
• Originality score distribution chart — rendered when the dataset includes any rows with non-null originality scores
• Confidence-category pie chart — rendered when the dataset includes GPTZero rows
• Flagged-submissions table with drill-down to the per-submission detail modal

Filtering the Report

• Date range — limit the report to a specific time window
• Group — limit to submissions from a specific group (Educator required)
• Assignment — limit to one assignment
• Provider — filter by which detection provider produced the result (useful when looking at historical data after a provider switch)
• Flagged only — show only submissions that crossed a threshold

Teacher data isolation applies. Teachers see only submissions from groups they own. Administrators see everything. The anonymous-grading mask also applies — submissions that haven’t been graded yet show “Submission #N” instead of student name.

What Students See

Nothing. Plagiarism results are private to graders and administrators by design. Students don’t see the panel in the grading interface (because they don’t have access to it), and the per-submission detail modal isn’t surfaced anywhere on student-facing pages.