Portfolio optimization faces challenges due to sensitivity in traditional mean-variance models. The Black-Litterman model mitigates this by integrating investor views, but defining these views remains difficult. This study explored the integration of large language models (LLMs) generated views into the Black-Litterman framework for portfolio optimization. Our method leverages LLMs to estimate expected stock returns from historical prices and company metadata, incorporating uncertainty via variance in predictions. We backtest optimized portfolios using LLM against the S&P 500 for 2024, assessing its returns and risks. Empirical results suggest that different LLMs exhibit varying levels of predictive optimism and confidence stability, which impact portfolio performance. The source code and data are available at https://anonymous.4open.science/r/ICLR2025-4E6D.