Methodology
Transparent signals, not a black-box verdict.
BareGPT analyzes common patterns in AI-assisted writing and highlights signals that influence the risk score. The goal is to help writers revise, not to declare authorship.
Signal categories
We combine several interpretable signals such as repetition, transitions, and sentence smoothness. Each signal is weighted to form the overall score.
What the score means
The score reflects likelihood of AI-like patterns, not proof. Use it as a diagnostic to improve drafts.
Best input length
Longer samples are more reliable. We recommend at least 150 words for stable results.
Limitations
AI detection is probabilistic. Human writing can look AI-like and AI can imitate human style. Always review manually.
Data handling and privacy
We process text to generate scores and reports. See the privacy policy for details on data retention and access.