AI text detectors can reliably identify code written by AI models.

Is it true that AI text detectors can reliably identify code written by AI models?

Partly. One peer-reviewed study reports a detector reaching AUC 0.87 on code from a single model, while other peer-reviewed work shows detectors struggling on mixed text and being evaded 78.6-96.2% of the time. None of the items establishes reliable code detection in general; the strongest code-specific result covers one model and one method.

Verdict: CONTESTED PARTIAL · 1 source for, 4 against · Last checked 3 Oct 2026

The private library was unavailable for some searches in this check. Missing library results are not evidence against the claim.

Xu, Z. & Sheng, V. S. (2024). Detecting AI-Generated Code Assignments Using Perplexity of Large Language Models. doi.org. https://doi.org/10.1609/aaai.v38i21.30361

Zhang, Q., Gao, C., Chen, D., Huang, Y., Huang, Y., Sun, Z., Zhang, S., Li, W., Fu, Z., Wan, Y., & Sun, L. (2024). LLM-as-a-Coauthor: Can Mixed Human-Written and Machine-Generated Text Be Detected?. doi.org. https://doi.org/10.18653/v1/2024.findings-naacl.29

David, I. & Gervais, A. (2025). AuthorMist: Evading AI Text Detectors with Reinforcement Learning. doi.org. https://doi.org/10.48550/arXiv.2503.08716

Lesnov, R. (2026). In 2026, Can AI "Humanizers" Outsmart AI Text Detectors?. doi.org. https://doi.org/10.2139/ssrn.7128878

Hadi, M., Anastasia, G., & L., O. D. (2026). Explainability-Based Token Replacement on LLM-Generated Text. zenodo.org. https://doi.org/10.5281/zenodo.20518113