Of all the young AI startups currently being relentlessly pursued by VCs, GPTZero achieved profitability in its first yr and a half, generating multi-million-dollar revenues. Founded by 24-year-old Edward Tian and 26-year-old Alex Cui, who have been friends since highschool, GPTZero offers a detection tool that helps determine whether content has been generated by artificial intelligence.
The founders decided to participate in a $10 million “preemptive” Series A round led by Footwork co-founder Nikhil Basu Trivedi, the team told TechCrunch exclusively. (“Preemptor” is a VC term for a situation where an investor enters into a deal before the founders have tried to boost funds.)
This is a real coup for Basu Trivedi. GPTZero has been under the radar of top VC firms virtually since Tian launched the first version of the web app in December 2022, and 30,000 people immediately flooded the app, crashing the Streamlit-hosted site. (Adrien Treuille, co-founder of Streamlit, who sold Snowflake for $800 million, later became an angel investor, Tian says). The company formally began operations in January 2023.
They say that throughout 2024, as their customer base grew, the young founders received 4 to 5 calls from VC investors per week.
GPTZero has seen a 500% increase in ARR over the last six months, the founders told TechCrunch, adding that its user base has grown from 1 million to 4 million over the last 12 months. This makes it one of the fastest growing consumer apps of the yr, by certain means.
They said the company has been profitable for the past few months, adding that it has more cash in the bank than has been raised in total over the life of the company. Number-wise: over $13 million from an initial $3.5 million to a new $10 million.
And the growth continues. User numbers and revenue “have more than doubled, maybe even tripled, since January,” Basu Trivedi said. While they didn’t comment on the valuation, based on a typical Series A round of 20%, the deal valued the company at roughly $50 million before money. Other investors in this round include education-focused (and women-led) Reach Capital; Jack Altman’s Alternative Capital; Uncork Capital (Jeff Clavier’s fund); and Neo (Ali Partovi Fund).
How the VC won the deal
Basu Trivedi, a Princeton graduate, took the lead in this deal by playing the long game. He met Tian in 2022, before the GPTZero craze, during an annual event where a small group of Princeton students visit Silicon Valley firms. Basu Trivedi at all times takes the group on a tour of Stanford Dish.
Tian developed GPTZero while studying computer science, natural language processing, and journalism at an Ivy League school. During his internships at the BBC and later at The New York Times, he wrote code that helped journalists discover content generated by artificial intelligence.
After the crazy response his first web app received, Tian turned to his buddy Cui for help. Cui earned a master’s degree in machine learning from the University of Toronto, but left the PhD program to change into a co-founder.
The two rewrote the app on its current standalone platform and raised $3.5 million in seed funding after reaching roughly 1.5 million users in the first five months. This got here mainly from angel investors reminiscent of Tom Glocer, former CEO of Reuters; Russ Salakhutdinov, professor at Carnegie Mellon University and former director of artificial intelligence research at Apple (after selling his startup Perceptual Machines to Apple in 2016); and Mark Thompson, CEO of CNN and former CEO of the New York Times.
Basu Trivedi has seen GPTZero gain press and impressive angel followings – and has heard rumors about it among the VC hatches. As a seed investor who has backed firms like Canva, ClassDojo, and Frame.io, he knew a company was hot when he saw one.
In January 2023, he texted Tian to ascertain in. He convinced the founders with his web and product knowledge from fast-growing firms like Canva and the experience of his fund’s co-founder, Mike Smith, former chief operating officer of Stitch Fix and Walmart.
The two founders, each in their 20s, “were looking for investors with experience in both product and operations, especially as Alex and I are learning how to build a large company,” Tian said.
To prove this point, shortly after the round closed, Footwork organized a networking event with AI leaders including Jack Altman (brother of OpenAI’s Sam Altman, who joined the Seed A round), who was also a college classmate of Basu Trivedi and the founder of Nvidia, CEO Jensen Huang.
“The advantage of big data“
GPTZero is not the only company involved in identifying artificial intelligence-generated content. Others include AI Writing Check, Copyleaks, GPT Radar, CatchGPT and Originality.ai.
However, many people in the AI detection industry have terrible accuracy, researchers find. So much so that OpenAI, which was pressured by AI industry paranoia to launch its own AI detector in early 2023, shut down the tool about seven months later in July after widespread criticism for its poor performance.
Interestingly, when TechCrunch’s Kyle Wiggers conducted his own experiment with these tools, all but GPTZero failed.
Of course, GPTZero has its own standards, especially because of its partnership with Penn State researchers, which help it show that its technology works well despite the industry’s overall fame.
Cui says GPTZero is more accurate because it has access to more data and has built its own LLM models using the most advanced open source tools, which it does not disclose.
“We have a big data advantage. We have millions of examples of human-versus-AI text,” Cui said. “We also combined this with best-in-class models and deep learning. We actually use language models to detect language models.”
While the startup could also be best known for helping teachers detect AI-generated student work (in October GPTZero has reached an agreement with the American Federation of Teachers) the customer base has grown. It now includes government procurement agencies, grant writing organizations, hiring managers, and – particularly interestingly – those responsible for labeling AI training data.
It seems that using AI-generated data to coach the AI ”causes the model to break down,” Tian says, because training the model from fabricated examples is not the best approach to ensure it functions in the real world.
Naturally, young founders have a more impressive, long-term vision. They need to create a new, independent layer of the Internet that will likely be accountable by ensuring correct attribution of human and artificial intelligence content.
To this end, the team is currently working on hallucination detection through artificial intelligence. Hallucinations, in which AI presents AI-generated fiction as if it were fact, are the bane of the GenAI industry. The company’s first step towards solving this problem is its newly available free AI text copyright checking feature for LLM training datasets. This will help them generate training data for broader hallucination detection.
“We’re just trying to avoid a world where the entire Internet is made up of AI-generated content,” Tian said. “An internet where everyone uses artificial intelligence does not provide opportunities for people to continue contributing creative and original content.”