The rise of convincingly human-like text generated by AI has opened a Pandora's Box of potential misuse, from sophisticated spam and propaganda to academic plagiarism. Distinguishing between human and machine-authored text is becoming increasingly challenging, a problem researchers are tackling with innovative “watermarking” techniques. But current methods, like the “red-green” watermark, are vulnerable to attacks like paraphrasing, which can easily erase the telltale signs embedded in AI-generated text. Researchers are now exploring “ensemble watermarks,” a more robust approach that combines multiple subtle markers to create a more resilient fingerprint. This technique draws inspiration from the field of stylometry, which analyzes linguistic patterns unique to individual writers. By weaving together various stylistic cues—like the use of specific sensorimotor words (related to sensory perception and actions), acrostic patterns in sentence beginnings, and the established red-green watermark—the ensemble approach makes it significantly harder to disguise AI-generated text. This research tested the ensemble method across different large language models (LLMs) and watermark strengths. The results show that combining these features dramatically increases the detection rate of AI-generated text, even after paraphrasing attempts. Notably, while individual components like acrostics might be weaker on their own, they contribute to the overall strength of the ensemble, offering a layered defense against manipulation. While the ensemble approach increases the complexity of detection algorithms, the same detection function can be used for all ensemble configurations, a significant advantage. This research represents a critical step in ensuring accountability and combating the potential misuse of LLMs. As AI-generated text becomes increasingly prevalent, robust watermarking techniques like the ensemble approach will be essential in maintaining trust and transparency in the digital world. Future research will likely explore additional stylometric features and more sophisticated sampling strategies to strengthen these digital watermarks further, paving the way for a more secure and accountable AI-driven future.