Quality Assurance
Under Construction
Here we will publish the results of public stress tests of our algorithm.
- Deception attempts: how the algorithm handles prompts commanding "human-like" style and artificial typos.
- False Positives breakdown: analysis of real texts where the detector made a mistake, and explaining the reasons.
- Complex hybrids: dissecting texts with deep authorial editing layered on top of AI drafts.
We believe in transparency. A tool should not only show its victories but also openly analyze its blind spots. The first batch of cases will appear here soon.