Pangram – AI detector for text and images

Pangram is an AI-detection service for both text and images that claims industry-leading accuracy and third-party verification from researchers at the University of Maryland and the University of Chicago. It uses proprietary NLP models and a large, diverse training corpus to identify structural, stylistic, and semantic patterns associated with outputs from major LLMs (OpenAI, Anthropic, Google, Meta, xAI, etc.), benchmarking about 26 models and touting overall accuracy around 99.7%. Key product capabilities include segment-level AI analysis, detection of AI-assisted edits versus fully generated passages, an integrated plagiarism checker, and the ability to detect “humanized” text that has been edited to evade detection. Image uploads (JPG/PNG/WebP, min 512×512, up to 30MB) are supported with a limited free-scans allowance.

The workflow supports paste/upload of common file types (PDF, DOCX, RTF), bulk scanning (up to 100 files), detailed reports with highlighted segments, and historical records. Integrations include a Chrome extension, Google Docs, Google Classroom, Canvas, and an API for high-volume use. Target customers span education, publishers, legal and HR teams, and platforms with user-generated content; testimonials and academic comparisons claim lower false-positive rates than competitors and even trained humans. Founded in 2023 by AI researchers from Tesla and Google, the company positions its mission as restoring trust and transparency in an era of widespread generative AI.