AI tools and text humanization technologies have become increasingly common in today’s digital landscape, reshaping how written content is produced, revised, and evaluated. This study investigates the effectiveness of AI “humanizer” tools in transforming AI-generated text and the ability of detection systems to identify such modifications. With the growing use of generative AI in writing, concerns have emerged regarding authenticity, academic integrity, and the reliability of AI-detection technologies, particularly in educational and research settings.

The research conducts a systematic comparison between AI-based text modification tools and widely used detection systems. It evaluates whether modified texts can bypass detection while maintaining clarity, coherence, and the original meaning of the content. In addition, the study explores how different levels of modification influence the balance between linguistic naturalness and detectability, highlighting potential trade-offs between evasion capability and content quality.

Using standardized AI-generated texts, the study applies both computational analysis and human evaluation to assess readability, naturalness, and semantic consistency. Statistical techniques are employed to examine relationships between detection scores, linguistic quality, and perceived human-likeness, providing a more comprehensive understanding of how these systems perform under varying conditions.

The findings are expected to offer valuable insights into the strengths and limitations of current AI-detection systems, particularly in academic contexts where integrity and transparency are critical. The study contributes to policy development, supports responsible AI use in education, and aligns with broader efforts to promote ethical, transparent, and accountable integration of AI technologies in digital writing environments.