Our take
Currently the strongest detector in independent testing we’ve reviewed. Still a screening tool, not a verdict.
At a glance
- Best for: Anyone who prioritizes a low false positive rate, including researchers and publishers
- Free option: Free checks available; see its site for current limits
- Website: pangram.com
Main features
- Sentence-level results
- Trained to catch paraphrased and humanized AI text
- API for developers and platforms
Strengths
- Best performer in the 2025 University of Chicago study, with near-zero false positives on medium and long passages
- Held up against a commercial humanizer in the same study
- Publishes and links to third-party evaluations
Limits and concerns
- Less established in classrooms than Turnitin or GPTZero
- Strong benchmark results still don’t make any score proof of misconduct
Remember: no detector score is proof that someone used AI. How accurate are AI detectors?
Sources
Last reviewed October 2026. Features and free limits change often; check the product’s site for current details. We have no affiliate relationship with any product reviewed here.