A recent survey conducted by the Schreibzentrum at Goethe University reveals a complex relationship between students and artificial intelligence in academic writing. The study, involving 4,048 participants, highlights that while most students utilize AI tools like ChatGPT or DeepL, their attitudes remain deeply ambivalent regarding the practice. According to the findings, 89% of students integrate AI into their academic workflows, primarily for brainstorming ideas, overcoming writer's block, or refining language and style. However, its use for analyzing research texts or drafting substantive content is less common, with only about one-third of respondents reporting such usage. The majority of students attempt to restrict AI use to tasks they deem justifiable, driven by an awareness that these systems are prone to errors and that over-reliance could erode essential skills. Despite the utility of these tools, concerns are widespread. Slightly more than half of the respondents feel they achieve less independently when using AI, and 45.5% express worry about the loss of critical thinking abilities. Conversely, 92.2% maintain a strong sense of responsibility for the texts they submit, and 78% view writing as a vital method for developing ideas and learning. The study identifies a significant tension between students' ethical reservations and their practical behaviors. While 10% reject AI entirely and 18% manage to use it responsibly as a sparring partner or tutor, 72% admit to being tempted to let AI take over entire tasks, effectively acting in a "ghost" role. This shift is driven by a combination of time pressure, grade anxiety, motivational struggles, and the addictive design of chatbots that encourage intensive engagement. Some students noted the difficulty of actively motivating themselves to think independently rather than delegating tasks to the machine. Nora Hoffmann, the study's lead author, notes that the current situation presents a favorable starting point because many students still retain a clear understanding of what is lost when tasks are fully delegated to artificial intelligence. However, she warns that this perception is likely to change as AI becomes more ubiquitous. Consequently, the study recommends that universities act immediately to address these challenges. The research concludes that AI has already become an inseparable part of academic writing, yet its application remains contested. Universities face the pressing challenge of guiding students to use these tools in ways that support rather than undermine their intellectual contributions. Hoffmann emphasizes that the goal is not to ban AI, but to foster a culture where students can use the technology critically and thoughtfully, ensuring it serves as a supplement to, rather than a replacement for, their own cognitive efforts.