Imagine a world where we could understand the pulse of any city, from New York's bustling traffic to the carbon footprint of Tokyo, with just a few keystrokes. Large Language Models (LLMs) are starting to make this a reality. New research explores how these powerful AI tools can unlock hidden insights about cities across the globe, even in places where traditional data collection is challenging or impossible. By querying LLMs about specific metrics like pollution levels, public transport effectiveness, or even crime rates, researchers discovered that these models hold a surprising amount of knowledge about urban environments. The study didn't stop at simply asking direct questions. They also devised innovative methods to extract both explicit features (like population density or economic activity) and implicit, less easily interpretable features from the LLM's internal representations. These LLM-derived features were then used to train machine learning models, boosting the accuracy of predictions on a wide range of urban characteristics. While LLMs don't replace traditional data collection, this research shows their potential to fill gaps, broaden the scope of urban studies, and offer valuable insights for data-driven decision-making in city planning and management. One of the most intriguing findings is that LLMs demonstrate varying degrees of knowledge across different cities and topics. They excel at providing information on well-documented cities and widely reported metrics, like COVID-19 cases in major US cities. However, the research also reveals how to detect when an LLM is less confident, often resorting to generic or inconsistent answers when faced with unfamiliar tasks. This ability to discern an LLM's confidence level is crucial for responsible use of these powerful tools. The research also delved into the nuances of extracting features from LLMs, exploring the impact of different prompt languages and model sizes. Interestingly, the study found that explicit feature extraction, where the LLM is specifically asked for relevant features, is often more effective and versatile than relying on the model's hidden representations. This highlights the importance of carefully crafted prompts to unlock the full potential of LLMs in urban research. This work opens exciting new avenues for understanding cities worldwide. It empowers researchers and policymakers with a new tool to analyze urban dynamics, address urban challenges, and ultimately build more sustainable and thriving urban environments.