Ever since, AI has been introduced, AI-generated AI-generated text, images, and everything in between are showing up everywhere; in news feeds, LinkedIn posts, academic papers, even courtroom filings. Pangram, a New York-based startup, just raised $9 million on the bet that people are going to need real tools to tell what’s human and what isn’t. And that bet is looking pretty smart right now. The Pangram AI detector is designed to identify AI-generated text and images with high accuracy.
Pangram AI Detector Secures $9 Million and Launches New Tools
The round was led by Menlo Ventures, with Haystack, ScOp, Script Capital, and Cadenza also chipping in. Alongside the raise, Pangram launched two things at once: Pangram 4, its next-generation AI text detector, and Pangram Image, a new tool built to catch AI-generated pictures. The image detector is still in research preview, but a wider release is expected soon. Tools like AI image generators have made it easier than ever to produce visuals that look completely real.
Here’s the part that’s genuinely impressive. Pangram says its new text model is over 99% accurate at catching AI-assisted writing, and it’s gotten noticeably better at spotting content that’s been run through “humanizer” tools designed to trick detectors. That’s no small feat, considering how much effort has gone into building those evasion tools in the first place.
Why Two Stanford Graduates Built Pangram

The company was started about two years ago by Max Spero and Bradley Emi, two Stanford AI and machine learning grads who watched the internet change almost overnight after ChatGPT launched. Spero has described what followed in blunt terms: an explosion of SEO spam, bots, and what he calls LLM-powered disinformation campaigns. So, they built something to push back.
How the Pangram AI Detector Works
The way Pangram works is kind of clever. The model was trained on tens of millions of real human documents. For each one, the team created what they call a “synthetic mirror”, basically the same topic, length, and tone, but written entirely by an AI model instead. By comparing the two side by side, over and over, the system learned to pick up on the small stylistic habits AI tends to fall into, without relying on watermarks or hidden metadata.
Spero’s team ran this test across outputs from both Claude vs ChatGPT to make sure the model wasn’t just learning one AI’s habits
Key Performance Highlights of Pangram:
- Text detection accuracy: 99%+
- Roughly 1 in 10,000 human documents mislabeled
- Subscription pricing starts at $20 a month
Why AI Content Detection Matters More Than Ever

The stakes here go beyond embarrassment. Earlier this year, the research archive arXiv rolled out a policy that can ban authors for a full year if their submissions show signs of unreviewed AI output, like leftover phrases such as “Would you like me to make any changes?” Lawyers have already been sanctioned for citing fake cases that ChatGPT simply made up. This isn’t a theoretical problem anymore.
Where the Pangram AI Detector Is Already Being Used
The Pangram AI detector is already being used by publishers, schools, recruiters, and online platforms. Pangram’s tools are already showing up in places people actually use. There’s a Chrome extension that labels AI content in real time across platforms like X, LinkedIn, Reddit, and Medium, complete with a feed health score. Substack recently integrated Pangram’s technology directly into its platform, and other customers reportedly include Quora, along with schools, publishers, and recruiters.
The Bigger Picture for AI Content Detection
Spero isn’t framing this as some kind of witch hunt against people who use AI. His concern is bigger than that. “If we do not actively discriminate in favor of human content,” he said, “then we’re just gonna see more and more AI, and it’s just going to drown out any human signal that we have.”
As AI-generated content continues to grow, the Pangram AI detector could become an essential tool for verifying authenticity online.
Frequently Asked Questions (FAQs)
What is the Pangram AI detector?
The Pangram AI detector is an AI-powered tool that identifies AI-generated text and images. It helps publishers, educators, recruiters, and businesses distinguish human-created content from AI-generated material.
Why did Pangram raise $9 million?
Pangram raised $9 million to expand its AI detection technology, launch Pangram 4, develop its AI image detector, and strengthen tools that combat AI-generated spam and misinformation.
What is Pangram 4?
Pangram 4 is the company’s latest AI text detection model, designed to identify AI-assisted writing with over 99% accuracy while improving detection of content modified by AI “humanizer” tools.
Who can use the Pangram AI detector?
The Pangram AI detector is designed for publishers, educational institutions, recruiters, legal professionals, businesses, and content platforms that need to verify the authenticity of digital content.
Does Pangram detect AI-generated images?
Yes. Pangram recently introduced Pangram Image, an AI image detection tool currently available in research preview, with a broader public release expected soon.