The study found that AI-generated messages effectively mimicked personal writing styles across all participant groups, regardless of their own experience with AI tools. The researchers did not disclose whether certain message types or scenarios triggered greater skepticism than others, nor have they published detailed breakdowns of which demographic groups showed the strongest disclosure penalties. The work builds on earlier research documenting poor human detection of AI text and prior findings that disclosure itself reduces trust in apologies and job applications.
For attorneys advising clients on AI use in business communications, employment matters, or client-facing work, the findings present a practical problem: undisclosed AI use may improve initial reception but carries reputational risk if discovered. The disclosure penalty suggests that transparency about AI authorship—whether in marketing emails, client correspondence, or internal communications—may be legally and strategically preferable to relying on undetected use. As AI integration accelerates in workplace and commercial settings, the gap between perceived authenticity and actual authorship will likely become a material issue in disputes over misrepresentation and good faith dealing.