Business

Grammarly Rebrands as Superhuman, Acquires AI Detection Startup GPTZero in Push to Verify Online Content

The $88 million deal integrates GPTZero's hallucination and AI screening tools into the former Grammarly platform, shifting focus from text generation to content verification.

Superhuman, the productivity platform formerly known as Grammarly, acquired artificial intelligence detection startup GPTZero in June, folding a tool built by two college friends into a software ecosystem used by tens of millions of writers. The deal valued GPTZero at more than $88 million, according to PitchBook data, delivering at least a 10x return on the $13.5 million the startup raised during its independent run.

The acquisition signals a strategic pivot for the combined company. Rather than solely helping users generate or polish text, Superhuman is now positioning itself to verify whether that text was written by a human or a machine. GPTZero’s roughly 30 employees have been integrated into a new “authenticity” team within Superhuman, led by founders Edward Tian and Alex Cui. Their technology suite—spanning AI detection, hallucination flagging, plagiarism checking, and writing replay—is now embedded across Superhuman Go, an AI assistant operating in more than 1 million applications and websites.

GPTZero generates $30 million in annual recurring revenue from a base of over 20 million registered users. Footwork, the venture firm that led the startup’s $10 million Series A round in June 2024, noted that the company operated profitably almost immediately after receiving institutional capital. The sole outside board member before the acquisition, Footwork cofounder Nikhil Basu Trivedi, characterized the product’s core appeal not strictly as policing AI use, but as satisfying a broad consumer appetite to understand the provenance of online content.

The origins of the technology trace back to December 2022, weeks after OpenAI released ChatGPT. Tian, then a computer science senior at Princeton University, built the initial detector over winter break and deployed it on a free Streamlit hosting page. The site crashed within a week under traffic from 30,000 users. That early viral adoption forced Tian to abandon plans for a fellowship with the New York Times visual investigations team—a role he had secured after stints investigating disinformation with BBC Eye and war crimes alongside open-source intelligence group Bellingcat.

Cui joined shortly after the launch. The two had been friends since high school in Toronto, bonding over Model United Nations and science fairs. Cui brought prior experience combating online manipulation. As a Caltech undergraduate in late 2017, following the widespread dissemination of propaganda during the 2016 U.S. presidential election, he helped build Open Mind, a browser extension designed to expose readers to challenging viewpoints. Despite presenting the project to Facebook’s misinformation team and Senate staff on Capitol Hill, Cui said the effort collapsed because the team failed to ship a scalable product fast enough. Determined not to repeat that mistake, he left his master’s research in autonomous vehicles at the University of Toronto to join GPTZero full-time as chief technology officer.

Since scaling beyond its academic roots, GPTZero has focused heavily on auditing institutions where AI-generated errors carry severe consequences. In January, the company’s investigators identified more than 100 fabricated citations in papers accepted at NeurIPS, one of the premier global conferences for artificial intelligence research. Several prominent researchers publicly apologized and revised their submissions. A subsequent scan of 300 papers submitted to the International Conference on Learning Representations (ICLR) uncovered more than 50 hallucinated references. ICLR subsequently became a paying client, using GPTZero to screen incoming research.

The startup has also targeted corporate communications. Through a blog series documenting AI errors, GPTZero published evidence of hallucinations in reports issued by Deloitte Australia, KPMG International, EY Canada, and PwC Middle East. The company independently verified three of those cases; a University of Sydney academic flagged the fourth. None of the Big Four accounting firms responded to requests for comment regarding the findings. Separately, GPTZero analyzed transcripts from the 81st session of the United Nations General Assembly and classified approximately one in five spoken words as originating from AI-generated speeches.

Tian credits his approach to authorship and verification to John McPhee, the Pulitzer Prize-winning writer and longtime Princeton journalism professor. Tian was the only computer science major in McPhee’s final creative writing seminar, held during the pandemic when the then-nearly-90-year-old author relied on his student to troubleshoot Zoom calls. Tian still visits McPhee twice a year. When Tian once showed the author an AI-generated imitation of his prose, McPhee rejected it entirely, telling him it lacked his style. McPhee’s insistence on rigorous drafting and editing shaped Tian’s view of AI detection as merely an initial diagnostic rather than a final judgment on writing quality.

The broader AI detection industry faces persistent technical challenges. Statistical classifiers frequently struggle to distinguish between text produced by large language models and writing by non-native English speakers, raising concerns about false positives in academic settings. Several competitors that launched during the same post-ChatGPT window have either shut down or pivoted away from pure detection as generative models become increasingly difficult to fingerprint. GPTZero’s transition into a larger productivity platform insulates it from the volatility of operating as a standalone screening tool.

Despite the eight-figure valuation and acquisition, the founders maintained austere personal habits. Following the close of the Superhuman deal, Basu Trivedi found Tian and Cui staying at a $100-per-night hostel-style hotel in San Francisco during their celebration dinner. Cui currently serves as director of engineering at Superhuman while retaining his CTO title. Neither NeurIPS, ICLR, nor the Big Four firms named in GPTZero’s audits have publicly disputed the company’s specific citation findings as of October 2026.