Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/338288 
Year of Publication: 
2025
Series/Report no.: 
CESifo Working Paper No. 11316
Version Description: 
This Version: March 2026
Publisher: 
Munich Society for the Promotion of Economic Research - CESifo GmbH, Munich
Abstract: 
Generative AI (Gen-AI) is increasingly used to communicate on behalf of organizational leaders. We investigate whether Gen-AI can mimic a CEO and whether employees display aversion to its communication. We present a framework distinguishing between a strong or weak form of Gen-AI aversion. In a field experiment, employees evaluated responses from their CEO ("Wade") and an AI bot trained on the CEO's communication. Employees correctly identified the source 59% of the time, yet rated responses believed were AI-generated as less helpful, regardless of actual source. A second study with a general audience directly manipulates AI labels and likewise finds AI-labeled responses are rated as less helpful. Together, the findings provide evidence of strong algorithm aversion to Gen-AI used as a CEO bot.
Subjects: 
Generative AI
algorithm aversion
CEO communication
JEL: 
O15
M54
Document Type: 
Working Paper
Appears in Collections:

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.