Key points are not available for this paper at this time.
Purpose The study aims to investigate the influence of HIX.AI, an artificial intelligence (AI) tool that humanizes the generated content, on the detection capabilities of AI-generated text detectors. Design/methodology/approach The study investigates the reliability of six AI-generated content detection tools by passing ten essays, five each generated using Chat Generative Pre-Trained Transformer (ChatGPT) and Bard (Gemini) before and after passing through HIX.AI, which humanizes the AI-generated content. Findings The study found that the selected AI-generated text detectors identified the generated content with inconsistencies. Some of the essays were falsely identified as human-written by a few detectors, indicating that the detectors are unreliable. Post-HIX.AI application found that all the essays were passed as human-written except two, which identified as AI-generated and mixed content by two separate detectors. Practical implications The findings present the evolving field of AI-generated text detectors and the tools that can bypass the detectors highlighting the difficulties in identifying the generated content in the presence of the humanization tool. Passing the generated content as human-written has serious consequences, especially in academics. Hence, the study recommends more robust detectors to distinguish human-written and AI-generated content accurately. Originality/value The study contributes to the existing literature on AI text detectors and highlights the challenges that humanization tools pose in identifying AI-generated text by AI text detectors.
Ayub et al. (Wed,) studied this question.