Free Pangram AI detector

Yes. Every check is sent to Pangram's own detection API (currently Pangram 4.1) and the verdict, percentages and section labels you see are Pangram's. Fire Pixel pays for the API usage and is independent of Pangram Labs.

Yes, with no account or email. Each visitor gets up to 10 checks or 3,000 words a day, with up to 1,000 words per check. The free checks share a daily budget, so on a busy day they can run out early. They reset at midnight UTC.

Pangram is built to detect text from the major large language models, including ChatGPT, Claude, Gemini and Llama, rather than one model in particular. It also flags text that has been run through AI paraphrasing or "humaniser" tools, though those are harder to call.

AI-assisted text reads as written by a person but edited, polished or partly rewritten by an AI tool. It sits between fully AI-generated and fully human, and the section-by-section view shows where it applies.

Independent studies rank Pangram among the most accurate AI detectors, with very few false positives on human writing. No detector is proof on its own: short passages, heavy editing, translation and humanised text are all harder to judge. Treat a result as evidence to ask questions about, not a verdict on a person.

A sentence or two does not carry enough signal for a reliable call, from any detector. Paste a full paragraph or more for a result worth acting on.

No. The text is sent to Pangram's API for analysis and the result comes back to your browser. Fire Pixel does not save, log or read it, and the page masks the text from session recording.

For longer documents, bulk checks or API access, use Pangram directly. The offer under the checker gives 20% off; it is a referral link, so Fire Pixel gets a small credit if you sign up.