Understanding Copyleaks AI Detection Results
When reviewing Copyleaks AI detection reports, it is important to understand what the results do and do not mean. AI detection tools can be helpful indicators, but they should be interpreted with care and used alongside other forms of academic judgment.
What the AI Percentage Represents
The AI percentage shown in Copyleaks is not a confidence score.
Instead, it indicates the portion of the document that Copyleaks has identified as likely AI‑generated text.
For example:
- A score of 100% does not mean the system is 100% certain the work was written by AI.
- It means that 100% of the document’s content has been flagged as likely AI‑generated by Copyleaks’ analysis.
Because of this, AI detection percentages should not be treated as definitive proof that a student used AI tools.
Why Some Clearly Human Text May Be Flagged
Copyleaks analyzes documents by dividing the text into larger sections (or “chunks”), rather than evaluating individual sentences in isolation.
- If enough content within a section is identified as likely AI‑generated, the entire section is marked as AI.
- This can result in legitimate human‑written text, such as institutional names, assignment titles, or standard phrasing, being flagged if it appears within a section that otherwise resembles AI‑generated writing.
This behaviour is expected and does not mean those specific phrases were generated by AI.
Understanding Sensitivity Levels
CopyLeaks offers three sensitivity levels that control how aggressively the system flags AI‑like text.
- Level 2 (Default: Recommended)
This setting provides the best balance and is designed to identify AI‑generated content that may have been lightly edited or modified by a human. It is the recommended setting for most instructional contexts. - Level 1 (Lower Sensitivity)
Produces fewer AI flags overall and is less likely to produce false positives. However, it is more likely to miss AI‑generated text, particularly if a student has revised or paraphrased it. - Level 3 (Higher Sensitivity)
Flags more content as AI‑generated and is more likely to identify subtle AI‑like patterns. This setting also increases the risk of false positives, including the flagging of human-written text.
Recommended Use in Instructional Decision‑Making
Copyleaks results should be used as one data point among many, not as standalone evidence of academic misconduct. Instructors are encouraged to also consider:
- The student’s writing history
- Draft submissions or revision processes
- In‑class work or assessments
- Direct conversations with the student
AI detection results should inform, but not solely determine, academic integrity decisions.
Learn More
For additional details on how Copyleaks AI detection works, including examples and limitations, please review Copyleaks’ AI detection documentation.
