> ## Documentation Index
> Fetch the complete documentation index at: https://docs.k16solutions.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Interpret results and FAQ

> Apply AI Detection results responsibly and resolve common reporting questions.

Use AI Detection to focus attention, then make decisions through informed human review.

## Interpretation principles

1. Confirm the reporting population and active filters.
2. Review the full submission, not only highlighted sentences.
3. Treat classification and confidence as model signals.
4. Consider assignment rules, writing context, and other evidence.
5. Follow your institution's documented process.
6. Give students the review and response opportunities required by policy.

<Warning>
  Do not create a local misconduct threshold from an AI percentage unless your institution has
  adopted a validated policy that explicitly requires it.
</Warning>

## Frequently asked questions

<AccordionGroup>
  <Accordion title="Does AI Only prove that a student used AI?">
    No. **AI Only** is the detector's classification for the processed text. It can justify closer
    review, but it does not establish authorship or intent.
  </Accordion>

  <Accordion title="What does the percentage measure?">
    A submission confidence or probability score describes the detector's assessment of that
    submission. Dashboard percentages can summarize a different population or metric. Read the
    displayed metric description before comparing them; a confidence score does not measure the
    exact share of text generated by AI.
  </Accordion>

  <Accordion title="Why are no sentences highlighted?">
    Sentence-level results use separate evaluations from the overall submission. Individual
    sentences may not be marked even when the complete text receives a nonzero overall score.
  </Accordion>

  <Accordion title="Why are many sentences highlighted when the score is not 100%?">
    Each sentence is evaluated individually, while the overall score summarizes the full submission.
    Many sentence signals can coexist with an overall result below 100%.
  </Accordion>

  <Accordion title="Why is a submission missing?">
    Confirm the course, term, date, and classification filters. The submission also must be
    available to the configured integration and eligible for processing. Record the course and
    submission details if it remains absent.
  </Accordion>

  <Accordion title="Why is a term missing from the dashboard?">
    The reporting dataset must contain eligible processed results for that term. Clear other filters
    and verify that relevant course submissions exist.
  </Accordion>

  <Accordion title="Why do two reports show different totals?">
    Compare filters, date boundaries, course scope, and whether the course report displays all
    submissions. The reports may summarize different populations.
  </Accordion>

  <Accordion title="Can I use the result by itself in an academic-integrity case?">
    K16 documentation does not recommend that. Use the result as one input within your institution's
    human review and evidence process.
  </Accordion>

  <Accordion title="Who can see AI Detection reports?">
    Access depends on the workspace's AI Detection configuration and reporting permissions. Ask your
    K16 administrator to review access when a person can or cannot open the dashboard.
  </Accordion>
</AccordionGroup>

## A practical review record

When your policy permits it, record:

* course and assignment
* submission and report date
* filters used to locate the result
* classification and overall score as displayed
* relevant highlighted passages
* contextual evidence reviewed
* reviewer and next step

Keep that record in your institution's approved system.

## When to ask for help

Contact your K16 administrator or support when:

* the workbook repeatedly fails to load
* expected processed submissions remain absent after you clear filters
* course and institution reports cannot be reconciled
* your access does not match your role

Include the workspace, course, submission identifier, time, active filters, and exact message. Do not send student work through an unapproved support channel.