<p>To the best of our knowledge, this is the first study to employ generative artificial intelligence (AI) to measure inflation expectations using online data. This novel approach overcomes key limitations of existing machine learning methods by accurately distinguishing between the perception of current/past inflation and actual forward-looking inflation expectations. Unlike conventional sentiment analysis, which struggles to differentiate these aspects, our method leverages a Large Language Model to extract inflation expectations directly from expert economic discourse on Twitter. This AI-driven approach offers several advantages: it enables real-time measurement of inflation expectations without costly and time-consuming surveys, enhances the timeliness and frequency of inflation expectation data, and provides policymakers, investors, and analysts with early insights into shifting expectations. By focusing on curated discussions among economists and financial analysts, our methodology reduces noise associated with broader public sentiment, improving accuracy and relevance. Moreover, the approach is highly scalable and can be expanded to incorporate additional data sources, including financial news and alternative social media platforms. The proposed indicator serves as a complementary measure to official inflation expectations indices, offering a more timely and flexible perspective that aligns with, but also extends beyond, conventional survey-based measures. Validation using data from multiple economies confirms its robustness across different economic contexts.</p>