AI technology transforms real estate valuation accuracy

Recent developments in artificial intelligence are reshaping real estate valuation by addressing the limitations of traditional Automated Valuation Models (AVMs). While standard algorithms excel at processing structured data such as square footage and bedroom counts, they historically struggle to interpret qualitative information found in prose.

A study published in the Journal of Real Estate Research utilized GPT models to extract quantitative features from residential property advertisements in Oslo. By integrating these extracted features into an XGBoost model alongside structured data, researchers improved root mean squared error by 24.31% and mean absolute percentage error by 15.25%.

Despite these advancements in data extraction, challenges remain regarding physical context. In high-stakes markets like Pacific Palisades, algorithms may fail to account for subjective or environmental variables such as property condition, views, and terrain. While platforms like Zillow and Redfin provide rapid estimates, the inability of AI to fully grasp non-standardized nuances can lead to significant pricing inaccuracies.