Pangram, an AI detection startup, raised $9 million on July 29 betting that demand for tools distinguishing human-generated content from AI-generated text will only grow.

The funding round was led by Menlo Ventures, with participation from Haystack, ScOp, Script Capital, and Cadenza. Pangram says its new text detection model is over 99% accurate at finding AI-assisted writing and mixed human-AI content—a critical capability as AI tools become ubiquitous.

The Detection Challenge

As generative AI models improve, spotting AI-written content becomes harder. Pangram’s model uses linguistic patterns and statistical signatures to identify text likely produced by large language models like ChatGPT or Claude.

The company focuses on enterprise use cases: publishers need to verify author claims, educators want to catch student submissions, and corporations need compliance tools for internal communications. Each market segment has different accuracy and latency requirements.

Market Timing

Pangram’s funding reflects broader market confidence in AI detection technology. LinkedIn recently launched its own AI-slop detection feature. Academic institutions have begun requiring AI disclosure. Publishers are building editorial policies around AI content.

The market for detection tools is nascent but growing fast. Pangram competes against both specialized startups and features being built into larger platforms. The company’s strategy focuses on accuracy and transparency—showing users why the model flagged content—rather than just binary classification.

The Longer Game

Detection may always lag behind generation. New models produce text humans struggle to distinguish from authentic writing. Pangram’s business depends on updating its detection algorithms faster than new AI models appear.

Pangram’s $9 million bet is a bet on authenticity. As AI-generated content floods the internet, proving you wrote something yourself may become as valuable as the writing itself.