Pangram

Pangram is an AI detection platform built by Pangram Labs, a Brooklyn-based company founded in 2023 by AI researchers from Tesla and Google. The tool detects AI-generated text and images, combining a proprietary deep learning model with plagiarism checking to give educators, publishers, and enterprises a complete picture of content authenticity. Pangram claims 99.98% accuracy and a false positive rate of 1 in 10,000, with those figures independently evaluated by researchers at the University of Chicago and the University of Maryland.

What It Is

Pangram is a web-based AI content detection service that analyzes text and images to determine whether they were authored by a human, generated by an AI model, or produced through AI assistance (partial AI editing). Unlike perplexity-based detectors, Pangram uses a neural network classifier trained on a diverse dataset of approximately 1 million documents, producing token-level embeddings and a classifier head that outputs a human/AI/AI-assisted prediction. The platform covers text from all major LLMs—including ChatGPT, Claude, Gemini, Grok, Llama, Mistral, and DeepSeek—and also offers an image detection capability currently in research preview.

Detection Architecture and Accuracy

Pangram's classifier tokenizes input text, converts tokens to embeddings, passes them through a neural network, and outputs a prediction with a confidence score. The model is trained on diverse datasets using hard negative mining and active learning rather than perplexity or burstiness metrics, which the company says reduces false positives on legitimate human writing. According to third-party research cited on the Pangram website from the University of Chicago Booth School of Business, Pangram's false positive rate across content types (product reviews, blog posts, news articles, novel excerpts, restaurant reviews, résumés) ranged from 0% to 0.8%, compared to higher rates for competing tools. The model undergoes continuous retraining as new AI models are released.

Key detection capabilities include:

• AI segment analysis showing which portions of a document are AI-generated
• AI assistance detection for text that has been partially edited by AI tools like Grammarly or QuillBot
• Detection of "humanized" text that has been processed to evade AI detectors
• Support for 25+ languages including English, Arabic, Chinese, French, German, Spanish, Japanese, and more

Update: Pangram 4

As of July 29, 2026, Pangram launched Pangram 4, described on the company blog as "the latest generation of our AI text detection model." A companion technical overview post details how Pangram 4 was built, evaluated, and how it compares to Pangram 3. Simultaneously, Pangram released a research preview of its AI image detection model, which accepts JPG, PNG, and WebP files (minimum 512×512px, up to 30MB). The blog also confirms detection of recently released models including GPT 5.6 and Claude Opus 5.

Deployment and Integrations

Pangram is accessible as a web application, a Chrome and Firefox browser extension, a Google Docs add-on, and a developer API. For educational institutions, it integrates directly into major learning management systems:

• Canvas
• Moodle
• Google Classroom
• Brightspace

The API supports usage-based pricing for developers and high-volume enterprise customers. File upload supports PDF, DOCX, and RTF formats, with batch processing of up to 100 files at a time.

Target Audience and Use Cases

Pangram serves a broad range of users across education, media, and enterprise:

• Educators and universities: LMS integrations allow teachers to run AI and plagiarism checks directly within their course shells; academic integrity departments use highlighted AI segments as evidence in policy enforcement.
• Publishers and media: Journalists and editors use Pangram to verify content authenticity before publication.
• Enterprise and compliance: Law firms, HR departments, and compliance teams use the API and enterprise license for high-volume content screening.
• Content moderation: Platforms with user-generated content use the API to reduce moderation workload.
• Developers and ML engineers: The API tier supports custom integrations and bulk processing with usage-based pricing.

Founding Team and Research Orientation

Pangram Labs was co-founded by Max Spero (CEO) and Bradley Emi (CTO), who met as Stanford undergraduates. Spero previously led active learning for autonomous vehicles at Nuro and held ML roles at Google, Two Sigma, and Yelp. Emi led the deep learning research group at Absci and was a member of the core computer vision team at Tesla Autopilot. Both hold B.S. and M.S. degrees in AI from Stanford. The company publishes research papers, maintains a model card, and collaborates with academic institutions on non-commercial research studies.