Faculty Mentor

Hayden Wimmer

Faculty Mentor Email Address

hwimmer@georgiasouthern.edu

Patterns in AI-Driven Misinformation: A Comparative Analysis of ChatGPT and Gemini

Location

Russell Union Ballroom

Type of Research

Proposed

I am a... (Select all that apply)

Graduate Student

Session Format

Poster Presentation

Select Your Campus

Statesboro Campus

College

Allen E. Paulson College of Engineering & Computing

Department

IT

Abstract

As Large Language Models (LLMs) become more widely used globally and shape public opinion, it becomes increasingly important to recognize their distinct tendency to produce, amplify, and disseminate misleading information. Preliminary findings indicate that, although these models are susceptible to hallucinations, their outputs exhibit variations in language sophistication and accuracy. Earlier studies have shown that LLM-generated misinformation is more difficult to identify than human-written content, while new comparative assessments have revealed significant disparities in accuracy, language style, and fact-checking skills. However, a research gap remains in systematically comparing the specific misinformation patterns between competing LLMs such as ChatGPT and Gemini, especially in real-world contexts. This research aims to address this by directly examining and contrasting their outputs to identify distinct trends and vulnerabilities while providing actionable insights for developing more resilient countermeasures to AI-driven misinformation by examining the implications for misinformation detection, governance, and digital literacy for the general public.

Program Description

.

Author Rights: Apply an Embargo

2-11-2026

Presentation Type and Release Option

Event

Start Date

4-23-2026 10:00 AM

End Date

4-23-2026 12:00 PM

This document is currently not available here.

Share

COinS
 
Apr 23rd, 10:00 AM Apr 23rd, 12:00 PM

Patterns in AI-Driven Misinformation: A Comparative Analysis of ChatGPT and Gemini

Russell Union Ballroom

As Large Language Models (LLMs) become more widely used globally and shape public opinion, it becomes increasingly important to recognize their distinct tendency to produce, amplify, and disseminate misleading information. Preliminary findings indicate that, although these models are susceptible to hallucinations, their outputs exhibit variations in language sophistication and accuracy. Earlier studies have shown that LLM-generated misinformation is more difficult to identify than human-written content, while new comparative assessments have revealed significant disparities in accuracy, language style, and fact-checking skills. However, a research gap remains in systematically comparing the specific misinformation patterns between competing LLMs such as ChatGPT and Gemini, especially in real-world contexts. This research aims to address this by directly examining and contrasting their outputs to identify distinct trends and vulnerabilities while providing actionable insights for developing more resilient countermeasures to AI-driven misinformation by examining the implications for misinformation detection, governance, and digital literacy for the general public.