AI
AI Detection Startup Pangram Raises $9M as Machine-Generated Content Overwhelms the Web
The New York firm just launched upgraded text and image detection models as demand grows for tools to identify AI-authored content across the internet.
TechCrunch — AI
A new funding round for Pangram, a startup dedicated to identifying machine-generated text and images online, underscores growing concern about the deluge of AI-created material flooding digital spaces. The company secured $9 million led by Menlo Ventures, with backing from Haystack, ScOp, Script Capital, and Cadenza.
The capital infusion accompanies the rollout of two detection tools: Pangram 4, a text classifier claiming over 99% accuracy at spotting AI-assisted or fully synthetic writing, and Pangram Image, a new visual detection model currently available as a research preview.
The Problem Pangram Aims to Solve#
Since ChatGPT's public debut, the internet has become a testing ground for machine-generated content — from SEO-optimized filler text to coordinated influence campaigns. Founders Max Spero and Bradley Emi, both Stanford-trained in AI and machine learning, started Pangram roughly two years ago in response to what they saw as an untenable flood of algorithmic output.
"I think it's just incredibly valuable to know whether what you're looking at is something that's AI-generated or not," Spero told TechCrunch. "Especially text that you're reading, because it changes how people approach the text."
The stakes have become concrete: lawyers have faced sanctions for citing fabricated cases generated by ChatGPT, academic repositories now ban authors who submit unreviewed machine output, and politicians have embarrassed themselves reading AI prompts on camera.
How the Detection Works#
Pangram's core technology trains a machine learning model on tens of millions of authentic human documents, then generates synthetic counterparts—identical in topic, length, and tone but written by frontier LLMs. By learning the consistent stylistic patterns of machine writing, the system identifies AI output without relying on watermarks or metadata artifacts.
The system goes beyond binary classification. It distinguishes between fully synthetic text and mixed human-AI content, recognizing scenarios where a person writes initially but uses AI for editing or refinement. This nuance matters because Spero believes disclosed AI assistance can be legitimate—transparency, not prohibition, is the goal.
Market Traction and Testing#
Pangram offers detection through multiple channels: a $20-monthly subscription with a Chrome extension that flags posts on X, LinkedIn, Substack, Reddit, and Medium, plus an API for institutional customers. to disclose which newsletters use AI assistance, and users include Quora, academic institutions, publishers, and recruiters.
