AI Detection Firm Pangram Faces Scrutiny Over Accuracy and Influence
Pangram, a small startup, has become a leading tool for detecting AI-generated text, with its scores leading to book cancellations and accusations against authors. Its integration into Substack and high-profile cases have raised its profile, but writers and experts question its reliability. The company's CEO defends its methods, but the debate over AI detection continues.
Pangram’s rise follows a pattern of rapid adoption despite limited funding—its $13 million total is a fraction of industry giants, yet its scoring has already influenced major publishing decisions, including Hachette’s cancellation of a novel and Substack’s integration of its tool. The company’s origins trace to cofounders with backgrounds at Google, Tesla, and biotech, and it initially launched as Checkfor.ai before rebranding. Its newest model, Pangram 4, arrived in July alongside a $9 million funding round, and the firm is expanding headcount by a quarter.
The company’s methods rely on AI to estimate AI involvement, producing percentage scores that have triggered public accusations against authors and outlets. Critics like author Jane Friedman describe widespread anger toward such software, viewing it as potentially more harmful than the AI it detects. Despite these concerns, Pangram’s CEO defends the technology, while the broader detection market—including rivals like Originality.ai and GPTZero—remains crowded and contested.
Pangram’s influence could reshape trust in written work across publishing, academia, and journalism, where a single score may determine careers or public reputation. Writers and researchers may face heightened scrutiny, while readers could over-rely on imperfect metrics. The company’s integration into platforms like Substack suggests such tools may become routine, potentially chilling creative expression or encouraging false accusations. Society may need clearer standards for what constitutes AI authorship, balancing innovation with fairness.