Georgia Institute of Technology
The Influence of AI-Like Text on Responses To Disclosure: Evidence From AI Detection Models
Abstract
dc:description.abstractThe rise of ChatGPT and other generative AI models has revolutionized machine-generated text. One potential application of this technology is helping craft firms’ narrative disclosures. Using two highly rated, commercially available AI detection models, I create novel measures of AI-like text in disclosure based on AI detection models’ classification that the text was generated either wholly or partly by AI. Using these measures, I study changes in disclosure surrounding the release of ChatGPT-4.0 in early 2023 and document a significant increase in the incidence of AI-like text in earnings conference call prepared remarks but not in managers’ responses to questions. Further evidence suggests that AI-like text in disclosure is more common among smaller, younger firms, and, on average, exhibits more positive tone, less uncertainty, and more forward-looking statements than non-AI-like disclosure text. I then compare the market responses to linguistic measures from AI-like disclosure text and non-AI-like disclosure text. Contrary to other studies that find generative AI text to be of higher quality and more persuasive than human text, my evidence suggests that tone in non-AI-like disclosure text is more strongly associated with returns. Overall, my results suggest that AI-like text may mute responses to information in disclosure.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Business
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Irons, Charles Austin
- Advisor dc:contributor.advisor
-
- Moon, Robbie
- Committee members dc:contributor.committeemember
-
- Blann, Justin
- Condie, Eric
- Shipman, Jonathan
- Yohn, Teri
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/78676