Landscape performance assessment is essential for evidence-based design, yet existing narrative evidence remains fragmented and difficult to synthesize. This study develops a Large Language Model (LLM)-assisted workflow to systematically examine 1,607 assessed benefits from 210 Case Study Briefs published by the Landscape Architecture Foundation (LAF). A binary framework was developed to appraise assessment quality and implemented using Gemini 2.5 Flash, demonstrating high consistency with expert validation. Results reveal a persistent hierarchy in assessment quality, with environmental benefits significantly outperforming social benefits. Longitudinal analysis indicates marked improvement in assessment quality between 2010 and 2020, followed by a quality plateau during the 2021-2025 period. This stagnation is associated with enduring limitations in empirical data collection, particularly the lack of primary data and direct measurements. The findings suggest that landscape performance assessment has reached the limits of report standardization and that future progress will require a shift toward technological integration, emphasizing advanced sensing technologies and validated measurement tools to strengthen empirical robustness.