Pangram Launches AI Detection API for Content Workflows

Summary

Pangram is offering an AI Detection API that lets developers add text analysis to products, moderation queues, learning-management submissions, content-management reviews, and other content pipelines. The service provides a realtime mode for individual decisions at $0.05 per 100 words, with a limit of five requests per second. Its bulk mode submits multiple documents as an ordered job and is priced 20% below realtime rates, with up to 1,000 billable units. Results use a consistent schema: overall fields include a prediction and short prediction, while fraction fields estimate the portions that appear AI-written, AI-assisted, or human-written. Window-level fields provide labels, scores, confidence, character ranges, word counts, and token lengths for analyzed sections. Pangram positions the API as a production decision layer, review-queue signal, or tool for scanning historical content at scale. Developers can create an account, generate an API key, install the Python SDK, and call `predict()` with text to analyze. Billing is based on prepaid credits from $5 to $2,000, with optional automatic refills. Enterprise customers can discuss SOC 2 compliance, zero-data-retention options, and higher rate limits; noncommercial academic projects may qualify for API credits. The page also cites use cases involving Quora and Gradpilot, without providing quantitative results.