AI Detector False Positive Scenario Calculator
A detector can flag genuine human writing. Explore how the share of submitted text that is actually AI-generated would affect the mix of flags in a hypothetical population. No essays are entered or uploaded; these numbers do not establish misconduct or authorship.
Work through the default example
Imagine 1,000 fictional texts, with exactly 100 classed as generated and 900 as human. Assume sensitivity of 80% for the first group and a false-positive rate of 5% for the second. The model produces 80 correctly flagged generated texts and 45 mistakenly flagged human texts, or 125 expected flags. 45 ÷ 125 = 36% of these hypothetical flags correspond to human text. That calculation is a group-average teaching illustration, not a measurement of any detector's current performance or a probability about an individual student's conduct.
Read our full guide to AI-writing detector limitations, including a discussion of published provider safeguards and the distinct role of human academic review. You can also use our Essay Revision Comparison Tool to compare your own genuine drafting history or the AI Claim Verification Tracker to document source checking. Neither tool certifies the origin of text.
Why this is not an authorship probability
Actual detector outputs may refer to the percentage of qualifying text categorized as likely AI-generated, not the share of submitted documents flagged across a population. Mixing these concepts produces incorrect conclusions. A real analysis would require validated performance rates on an appropriate test set and the actual sampling context; course policies and other evidence remain essential. Turnitin explicitly cautions that its report can misidentify human and AI-produced work and should not be used as the sole basis for adverse action.