JALT2025 - Tokyo

Efficacy of AI-text detection tools

Nov 1, 2025, 1:55 PM

30m

National Olympics Memorial Youth Center / 国立オリンピック記念青少年総合センター

Research-oriented Presentation (30-minutes) Technology Room 506

Speaker

MA. RITA ARANDA (National University, Philippines)

Description

The use of generative AI in academic writing raises concerns about authorship authenticity, especially regarding AI-edited content (Liu et al., 2024; Walters, 2023). This study analyzed 450 essays using a cross-sectional quantitative approach to evaluate detection accuracy. Enhancing training datasets and refining algorithms could improve detection tools for complex academic writing. Results showed that ZeroGPT outperforms SciSpace in detecting student-produced and AI-generated texts, demonstrating high reliability. Refining algorithms and training datasets could enhance detection accuracy.
Keywords: Artificial intelligence, detection tools, academic integrity, coherence, authenticity

Summary

The use of generative AI in academic writing raises concerns about authorship authenticity, especially regarding AI-edited content (Liu et al., 2024; Walters, 2023). This study analyzed 450 essays using a cross-sectional quantitative approach to evaluate detection accuracy. Enhancing training datasets and refining algorithms could improve detection tools for complex academic writing. Results showed that ZeroGPT outperforms SciSpace in detecting student-produced and AI-generated texts, demonstrating high reliability. Refining algorithms and training datasets could enhance detection accuracy.