Requirements engineering, the meticulous process of defining and managing system needs, is crucial for successful projects. But it's also complex and error-prone. Could AI help? Researchers are exploring whether Large Language Models (LLMs) can automatically verify that technical system specifications actually meet their requirements. Imagine giving an LLM a system blueprint and a list of requirements, and it tells you exactly where the blueprint falls short. This research puts that idea to the test, comparing LLMs against traditional rule-based systems in the smart grid domain. They looked at different factors like the complexity of the system, the number of requirements, and even the way they prompted the LLMs. The results? Advanced LLMs like GPT-4 and Claude 3.5 showed real promise, accurately identifying non-fulfilled requirements with impressive f1-scores between 79% and 94%. While not perfect, it hints at the potential of LLMs to streamline requirements verification, especially in the early stages of system design where catching errors is critical. Interestingly, the research revealed that LLMs perform better with more complex systems and struggle when faced with a long list of requirements all at once. This suggests that breaking down the task into smaller chunks could significantly boost accuracy. The way you talk to the AI matters too. Clear, structured prompting, along with a few real-world examples, helps LLMs deliver more reliable results. This research is a stepping stone towards more automated and efficient requirements engineering. Future work could explore how to mitigate the remaining inaccuracies and investigate the practical benefits of LLM-assisted system verification in diverse real-world projects.