Detect AI-Generated
Writing with Confidence
Advanced Transformer-based analysis with sentence-level precision. Built for educators who need reliable, transparent results.
Disclaimer: The percentage shown represents the probability that text may have been AI-generated. It should not be used as the sole basis for taking disciplinary action.
Submit Text for Analysis
Paste your text here, or drag & drop a file
Minimum 50 words · English text only · MVP
How Detection Works
A multi-stage pipeline inspired by state-of-the-art AI detection research.
Text Preprocessing
Clean, normalize, and split input text into sentences with precise character offsets for accurate highlighting.
Sliding Window Analysis
Combine adjacent sentences into overlapping windows of 5-10 sentences, enabling context-aware detection of AI patterns.
Probability Aggregation
Compute Gaussian-weighted averages across overlapping windows to determine each sentence's AI generation probability.
Threshold Protection
Apply conservative thresholds (p > 0.65) and the 20% masking rule to minimize false positives and protect students.
Built on Proven Technology
Drawing from the latest advances in AI text detection research and industry best practices.
Transformer Architecture
Deep contextual embeddings capture long-range statistical dependencies beyond simple perplexity and burstiness metrics.
Low False Positive Rate
Conservative thresholds and the 20% masking rule ensure results you can trust, protecting against misidentification.
Sentence-Level Precision
Every sentence receives an individual probability score with precise character offsets for accurate visual highlighting.
Designed for Academic Integrity
Our detection engine prioritizes low false positive rates above all else. In educational contexts, the cost of a false accusation is far greater than a missed detection. Results should always be used as a starting point for investigation, never as the sole basis for disciplinary action.