Related Experiment Video
Updated: Jun 4, 2026
07:14
Virtual Agent for Real-Time Motivational Interviewing by Integrating Adaptive Nonverbal Behavior and Language Models
Published on: December 23, 2025
How does ChatGPT respond to stuttering-related frequently asked questions? A mixed-methods, cross-version comparison
Amir Hossein Rasoli Jokar1, Hamid Karimi2
1Department of Communicative Sciences and Disorders, Michigan State University, United States.
Journal of Fluency Disorders
|June 2, 2026
Summary
Two ChatGPT versions were compared for stuttering FAQs. Both provided similar information, but ChatGPT-4.5 was more concise and structured, though slightly less readable. Clinician oversight is crucial for AI-generated health content.
Area of Science:
- Artificial Intelligence in Speech-Language Pathology
- Natural Language Processing for Health Communication
- Stuttering Research and AI Applications
Background:
- Frequently Asked Questions (FAQs) are a common source of health information.
- Large Language Models (LLMs) like ChatGPT are increasingly used for information retrieval.
- The accuracy and utility of AI-generated responses for specific conditions like stuttering require evaluation.
Purpose of the Study:
- To compare responses to stuttering-related FAQs generated by ChatGPT-4 and ChatGPT-4.5.
- To evaluate content accuracy, emotional tone, structural organization, and readability of AI responses.
- To assess the suitability of AI-generated content for public health communication regarding stuttering.
Main Methods:
- Developed 34 stuttering-related FAQs, refined by adults who stutter.
- Inputted each FAQ into ChatGPT-4 and ChatGPT-4.5 for response generation.
- Analyzed responses using reflexive thematic analysis, computational emotional analysis, and readability metrics.
Main Results:
- Both ChatGPT versions produced highly overlapping thematic content.
- ChatGPT-4.5 generated shorter, more structured responses; ChatGPT-4 provided longer, narrative explanations.
- GPT-4.5 responses were modestly less readable despite increased concision; emotional profiles were congruent.
Conclusions:
- ChatGPT serves as a basic explainer for stuttering FAQs, offering generally accurate, non-stigmatizing information.
- AI model iteration primarily impacts information framing, not core content.
- Clinician oversight, accessibility, and ethical considerations are vital for AI in stuttering public health communication.