Max Spero, chief executive of Pangram, is leading efforts to distinguish human-written content from that generated by artificial intelligence (AI). Pangram, a New York-based startup, has developed detection tools designed to identify whether a text—such as an article or social media post—was written by a human, an AI, or with AI assistance. As AI-generated content becomes increasingly widespread, Pangram aims to provide clarity amid concerns about distinguishing between human and machine authorship.
Founded in 2023 by Spero and Bradley Emi, both former Stanford computer science students with experience at Google and Tesla, the company was formed in response to the surge of AI-generated content following the introduction of ChatGPT in 2022. According to recent analysis by the digital marketing firm Graphite, approximately half of news articles and blogs today are AI-generated, underscoring the relevance of Pangram’s mission.
The company’s detection model, recently updated with the release of Pangram 4.0, operates through a user-friendly interface allowing individuals and organizations to input text and receive an assessment classifying it as human, AI, or a combination of the two. This model was trained on millions of examples to refine its accuracy. A 2025 study by the University of Chicago found Pangram’s tool achieved near-perfect accuracy for texts exceeding 500 words, outperforming similar services such as Originality.ai and GPTZero. However, the tool is less effective for shorter texts or those partially generated by AI—areas where the latest model seeks improvement.
Pangram’s client base includes individual subscribers, universities, publishers, and online platforms. Subscriptions are priced at $20 per month, and the company has secured partnerships with prominent platforms like Quora and Substack. Substack recently integrated Pangram’s technology to give readers the option to filter out AI-generated content. These developments have contributed to rapid growth in usage, with monthly users rising to 120,000 in June from 2,700 a year earlier. Pangram disclosed its annual revenue rate has increased about 35-fold during that period but did not provide specific figures.
Expanding beyond text, Pangram has introduced an AI-image detection model that it claims is 99.5 percent accurate at identifying fully AI-generated images. The company also plans to launch a tool for detecting AI-generated video later this year, aligning with its goal to provide users with what it calls “cognitive security”—a safeguard against fabricated or machine-generated content.
Despite its advances, Pangram acknowledges limitations in its technology. The tool can misclassify partially AI-generated text, which raises concerns about potential false accusations. Writer Freddie deBoer reported that appending a few AI-generated sentences onto his own text caused the entire passage to be flagged as AI-written. Spero responded that the new model addresses such scenarios with improved nuance. He also noted that while he personally refrains from using AI assistance in writing, he considers it acceptable for others to use AI as a tool. How institutions ultimately decide to handle texts partially produced by AI remains an open question.
Investors remain optimistic about the company’s prospects. Deedy Das, a partner at Menlo Ventures, which led Pangram’s latest $9 million funding round, described AI detection as a widely needed capability across industries. As AI-generated content becomes a norm rather than an exception, tools like Pangram’s may play an increasingly significant role in shaping how digital communication is evaluated and trusted.