Pangram – AI Detector

Pangram, an AI detection company founded in Brooklyn, NY in 2023 by AI researchers from Tesla and Google, has launched Pangram 4, its most advanced model, which detects AI-generated content with 99.98% accuracy. The tool, verified by researchers at the University of Maryland and the University of Chicago, supports over 20 languages and can detect content from major language models including ChatGPT, Claude, Gemini, Llama, and Grok, even after 'humanization' attempts.

We’ve launched our most advanced model yet: Pangram 4 Read about it here. https://www.pangram.com/blog/introducing-pangram-4 Detect AI-generated content with 99.98% accuracy. Trusted by universities, schools, and enterprises worldwide. Proven the most reliable and accurate AI detector on the market by third party researchers /blog/third-party-pangram-evals , including the University of Maryland and the University of Chicago. AI writing contains structural, stylistic, and semantic patterns that can all be detected by a trained eye or sufficiently sophisticated detection software. AI detection is not magic. It can actually work. Pangram's AI detector uses natural language processing and a massive data set of human and AI writing to analyze patterns in AI generated text created by popular models like ChatGPT, Gemini, Grok, Llama and Claude. Pangram's accuracy leads the industry and has been verified by researchers /blog/third-party-pangram-evals at the University of Chicago and University of Maryland. In an era where AI-generated content is increasingly prevalent in academia, media, and business, the ability to distinguish between human and machine-authored text is critical. In education, it creates trust between students and teachers. It's important in journalism where integrity is essential. AI-generated text isn't inherently bad. However, society can benefit from being able to distinguish between AI and human text. Our product Pangram combines cutting-edge AI and comprehensive plagiarism detection to give you the complete picture of text authenticity and get the information you need, all in one place. Pangram won't just tell you how confident it is that the document is AI, but also how much of it is. Pangram will highlight segments that have been edited by AI rather than fully AI-generated. Check for plagiarism and AI all at once. Perform an AI plagiarism check with one click. Features There are many AI detectors available, but Pangram is different. It actually works. We also provide advanced features backed by peer-reviewed research that allow users to understand the origin of a piece of writing. Uses proprietary technology with years of research and development behind it, not an open source model or commercial LLMs covered with branding. Pangram relies on diverse datasets, hard negative mining, and active learning to achieve industry-leading FPRs rather than perplexity and burstiness, which often fail in practice. Pangram can detect content from all major language models including ChatGPT, Claude, Gemini, Llama and more, making it a comprehensive solution for AI detection. Our AI detection works across more than 20 languages, making it a truly global, multilingual solution for institutions and businesses worldwide. Independent studies have shown that Pangram's AI detection outperforms trained human readers in identifying AI-generated content. Pangram can detect AI-generated text even after it has been "humanized," or processed by tools that attempt to evade AI detection, ensuring reliable detection. Our accuracy and reliability have been independently verified /blog/third-party-pangram-evals by third-party researchers and educational institutions. Who we are At Pangram, our mission is to rebuild trust amidst the increase in generative AI content. Founded in Brooklyn, NY, in 2023 by AI researchers from Tesla and Google, we continue to contribute to the research community to increase transparency. If you're interested, read more /blog/how-does-ai-detection-work about our contributions and collaborations. How it works Copy and paste your text into the field or upload files. Pangram supports PDF, DOCX, RTF files and up to 100 files at a time. Pangram analyzes your text using advanced AI detection algorithms to identify AI-generated content with high accuracy. Review detailed results showing AI detection scores, highlighted sections, and comprehensive analysis of your text. Access your complete history of scans, download reports, and manage your past detection records all in one place. Copy and paste your text into the field or upload files. Pangram supports PDF, DOCX, RTF files and up to 100 files at a time. Pangram analyzes your text using advanced AI detection algorithms to identify AI-generated content with high accuracy. Review detailed results showing AI detection scores, highlighted sections, and comprehensive analysis of your text. Access your complete history of scans, download reports, and manage your past detection records all in one place. Solutions Pangram's transparency tool reliably detects AI content and checks for plagiarism to help teachers understand the origin of written text and maintain integrity and fairness in assessment. Pangram's LMS integrations allow universities to create informed AI policies. Our AI detector's tool accuracy gives them the peace of mind to implement those policies at scale. Features 01 LMS Integration Our AI detector works with the most popular LMS systems making it easy to use across multiple platforms like Canvas, Moodle, Google Classroom, Brightspace, and more. Learn More → 03 API Solution Our API is perfect for high-volume projects with different pricing options to suit your needs. We have custom API solutions for educators, developers, and enterprise. Learn More → 02 Google Doc & Browser Extension Our browser extension is a quick and convenient way to use our AI detection tool while browsing the web or within web-based applications such as Google Docs. Download Now → See what our users say Alex Imas, Professor Pangram is pretty incredible. Much better than other AI detection tools by a country mile . I've used it and tested it thoroughly and seen minimal false positives and negatives. University of Chicago Booth School of Business For me, Pangram Labs has become the most student-centered option I’ve found for my courses. Students see the same report I see, directly in our Canvas shell, and Pangram’s documented false-positive rate 1 in 10,000 allows me to comfortably treat a flag as an invitation to talk, never as a presumption of guilt. — Susan Ray, Associate Professor Of English, Delaware County Community College There's a lot of overhyped, me-too products flooding the AI space, but this detector is almost supernaturally good. I have yet to experience a single false positive or false negative. Cloud Architect — Ryan Nicolace, Cloud Architect Jenna Russell, PhD We compare human experts with five AI detectors, including Pangram and GPTZero. Among automatic detectors, Pangram significantly outperforms the rest. Tuhin Chakrabarty, Professor Pangram is really one of the best AI detectors we have, and y'all should use it. Based on our experiences in the fall term, we have high confidence in Pangram as a detector of AI content in Wikipedia articles. — LiAnna Davis, Chief Programs Officer and Deputy Director at Wiki Education Pangram noticeably outperforms other LLM content detection services I've tried. It reduces our moderation workload enough to pay for itself many times over. If you run a platform with user-generated content, you should probably be using it. — Robert Mushkatblat, CTO LessWrong Dan Simon, CEO of Qwoted The future of journalism depends on trust. This is why we're delighted to be partnering with Pangram, which has set the gold standard for Al detection and attribution Our model How does Pangram achieve these results? Learn more. /research/how-it-works Our classifier uses a traditional language model architecture. It receives input text and tokenizes it. Then, the model turns each token into an embedding, which is a vector of numbers representing the meaning of each token. The input is passed through the neural network, producing an output vector. A classifier head transforms the output vector into a prediction of human, AI, or AI-assisted. We train an initial model on a small but diverse dataset of approximately 1 million documents composed of publicly licensed human-written text. The dataset also includes AI-generated text produced by GPT-5 and other frontier language models. The result of training is a neural network capable of reliably predicting whether text was authored by human or AI. After the initial training, our model undergoes continuous improvement through iterative refinement. We collect feedback from real-world usage and incorporate new data patterns to enhance detection accuracy. This iterative process ensures our AI detection remains at the forefront of technology, adapting to new language models and writing styles as they emerge. To maintain the highest level of accuracy, we consistently retrain our AI detector models with updated datasets that include the latest AI-generated content. This regular retraining process ensures our detection capabilities stay current with evolving AI technologies and can reliably identify content from the newest language models. Our classifier uses a traditional language model architecture. It receives input text and tokenizes it. Then, the model turns each token into an embedding, which is a vector of numbers representing the meaning of each token. The input is passed through the neural network, producing an output vector. A classifier head transforms the output vector into a prediction of human, AI, or AI-assisted. We train an initial model on a small but diverse dataset of approximately 1 million documents composed of publicly licensed human-written text. The dataset also includes AI-generated text produced by GPT-5 and other frontier language models. The result of training is a neural network capable of reliably predicting whether text was authored by human or AI. Third Party Evaluation Based on independent research from the University of Chicago Booth School of Business. Learn more. https://www.chicagobooth.edu/review/do-ai-detectors-work-well-enough-trust :~:text=The%20researchers%20tested Pangram 0.5% GPTZero 2.4% Originality.ai 1.7% RoBERTa 30.6% False positive rate — Product review Pangram 0% GPTZero 0.5% Originality.ai 0% RoBERTa 64.9% False positive rate — Blog post Pangram 0.1% GPTZero 1% Originality.ai 0.3% RoBERTa 77.8% False positive rate — News article Pangram 0% GPTZero 0.5% Originality.ai 0.3% RoBERTa 53.2% False positive rate — Novel excerpt Pangram 0.8% GPTZero 1% Originality.ai 2.2% RoBERTa 51.7% False positive rate — Restaurant review Pangram 0% GPTZero 0% Originality.ai 0.1% RoBERTa 69.1% False positive rate — Résumé Pangram 0.5% GPTZero 2.4% Originality.ai 1.7% RoBERTa 30.6% False positive rate — Product review FAQs Looking for answers? Explore common questions and get the information you need, all in one place.