ChatGPT’s Invisible Influence: The Silent Proliferation of AI-Linked Legal Documents Across U.S. Courts and Government Agencies

A recent study has revealed the presence of ChatGPT’s link-tracking tag in 298 federal court cases across 71 U.S. courts and in 133 filings with the U.S. Securities and Exchange Commission (SEC). Conducted by HumanizeMy.ai, the investigation traced the tag originating from ChatGPT hyperlinks. When a user copies a link given by ChatGPT, it includes “utm_source=chatgpt.com,” although users can easily delete this if they wish.

The presence of the tag in legal and regulatory documents suggests a notable reliance on ChatGPT by legal professionals, either for research or drafting purposes. The RECAP archive, a resource for federal court filings maintained by the Free Law Project, was the primary tool for detecting these tags in court documents. Notably, within the federal cases found, a significant concentration was observed in the Southern District of New York, the Northern District of California, and the Central District of California.

The cases that prominently exhibited these tags were mostly related to copyright, statutory causes, civil rights, personal injury, antitrust, and patent law. At the SEC, roughly half of the affected filings related to fundraising documents. Notably, one in five of the tagged links was directed towards market-research vendors, indicating ChatGPT’s employment in market size evaluations.

Beyond legal documents, the study uncovered the tag in several government documents, including materials from the U.S. Federal Register and records from congressional activities, as well as in documents from the U.K. government.

Firat Mihci, the founder of HumanizeMy.ai, emphasized that the study demonstrates only where a link from ChatGPT was copied, not the validity of the linked content or whether any text was directly authored by AI. He noted, “The tag does not prove anyone did anything wrong. It proves nobody looked.” This sentiment is echoed by industry figures like Oliver Roberts, of Wickard.ai, who acknowledges the tag as a “red flag” but not definitive proof of misconduct.

The findings spotlight an underlying question about the unnoticed and possibly widespread use of AI tools like ChatGPT in legal practices, particularly as regulators and legal institutions attempt to grapple with the ethical and procedural implications of AI in formal legal documentation.

The comprehensive report is available for download from HumanizeMy.ai.