AI content detection, like Pangram's AI transparency tool, analyzes a submitted document for traces of AI writing. AI detectors return a prediction and a score based on the likelihood that the submitted document is AI-generated.

Pangram's AI transparency tool analyzes a submitted document for traces of AI writing. If our system finds evidence that the document contains AI writing, it will return this observation along with a confidence number. Otherwise, we consider the result human-written.

Pangram is over 99% accurate according to multiple independent studies. Researchers from University of Maryland, University of Pennsylvania, and University of Houston have all published research showing that Pangram is over 99% accurate in their tests.

Pangram's false positive rate, the rate at which human documents are incorrectly flagged as AI, is currently 1 in 10,000. This number is calculated on an aggregate of public datasets containing tens of millions of written documents.

Yes! All user data is encrypted and stored securely on our servers to be displayed in your history. We never share or sell your data to any third parties.

Pangram detects AI writing in English, Arabic, Chinese, Czech, Dutch, French, German, Greek, Hindi, Hungarian, Italian, Japanese, Korean, Persian, Polish, Portuguese, Romanian, Russian, Spanish, Swedish, Turkish, Ukrainian, Urdu, Vietnamese, and more. For full details, see our announcement for our multilingual model release.

Pangram detects AI writing from ChatGPT, GPT-4o, GPT-4.5, Claude Sonnet by Anthropic, Google Gemini, Deepseek, Meta AI, Grok, Llama, Mistral, OpenAI's o1, o3, and Deep Research. Pangram also detects AI suggestions from AI writing tools like Grammarly and Quillbot. Pangram even detects new and unreleased AI models.

Pangram uses writing style, word choice, syntax choice, and grammatical structure to detect AI writing. Pangram only looks at the content of the text submitted. It does not compare the text to other submissions in a database or on the web like traditional plagiarism checkers, and it does not capture writing process data such as cut-and-paste. Pangram does not access OpenAI's database or analyze a user's browsing history.

Schools incorporate AI detection tools from Pangram by using LMS integrations like our Canvas integration or our Google Classroom integration. By bringing Pangram into the classroom, teachers can set boundaries around AI use and ensure students are improving their critical thinking skills through their schoolwork.

Professors use Pangram to set boundaries on AI use and ensure that students properly learn the material by using Pangram directly in their LMS. Academic integrity departments use Pangram to provide evidence for cases that goes beyond the score: highlighting the most common phrases that LLMs use and comparing submissions to actual AI-generated responses.

Many other AI detectors use metrics like perplexity to detect AI writing. Perplexity is how unexpected, or surprising, each word in a piece of text is, when looked at from the perspective of a particular language model or LLM. AI text is often low-perplexity, because of how AI models work. Unfortunately, other simple or unsurprising writing, like the Declaration of Independence, come up as a false positive, making perplexity-based detectors unreliable for scenarios like the classroom where false positives are undesirable. Pangram does not use perplexity or other metrics that correlate with AI writing, instead using a comprehensive deep learning approach that discovers patterns over millions of writing samples.

Many other AI detectors don't keep up-to-date with the latest AI models. Pangram is constantly collecting writing samples from the latest and most capable AI models to ensure that the software can detect writing from all AI services, even those from unreleased models or AI editing tools.

Pangram is built by a team of seasoned machine learning engineers with deep experience in AI. Pangram is the most advanced AI detection model and is built on millions of human and AI documents to ensure that it can pick up on even the most subtle nuances in AI writing.