The field of natural language processing is witnessing significant advancements with the rapid improvement of large language models (LLMs). A notable direction of research is the characterization and detection of texts generated by LLMs versus human-written texts. Studies have shown that human-written texts tend to exhibit simpler syntactic structures and more diverse semantic content, while newer models are producing texts with similar variability, pointing to an homogenization of machine-generated texts. Another area of focus is the development of reliable detectors for LLM-generated content, particularly in the context of web content where LLMs can produce unreliable and unethical material. Researchers are also exploring ways to evaluate and improve the performance of AI text detectors, including the use of few-shot prompting and chain-of-thought reasoning. Furthermore, there is a growing interest in adapting LLMs to specific stylistic characteristics, such as brand voice or authorial tones, to enhance enterprise communication. Noteworthy papers in this area include: Linguistic and Embedding-Based Profiling of Texts generated by Humans and Large Language Models, which reveals that human-written texts exhibit simpler syntactic structures and more diverse semantic content. Preprint: Did I Just Browse A Website Written by LLMs? proposes a reliable pipeline for classifying entire websites as LLM-dominant content. Beyond Easy Wins: A Text Hardness-Aware Benchmark for LLM-generated Text Detection presents a novel evaluation paradigm for AI text detectors that prioritizes real-world and equitable assessment.