Apple Music's new AI search cuts no-result rate by 86%

Apple Music serves over 150 storefronts in dozens of languages with a catalog growing by hundreds of thousands of new tracks daily. To improve search recall on misspelled, transliterated, and cross-lingual queries—especially tail queries that dominate unique searches—Apple researchers developed a multilingual semantic retrieval system. The core is a 305M-parameter Siamese bi-encoder fine-tuned from GTE-multilingual-base using curriculum-scheduled multi-objective training. It's integrated via a hybrid retrieval architecture that blends dense nearest-neighbor results with the existing token-based index using quantile distribution matching, allowing deployment without retraining downstream rankers. Offline, the model delivers a 69% relative improvement in Hit@10 over the base model.

In a worldwide online A/B test, the system produced a 2.28% relative conversion-rate lift overall and an 86% reduction in the no-result rate, with gains across every storefront and no observed regressions. The improvement is concentrated where it matters most: tail queries see a 7.93% relative CR lift, compared with 0.89% for mid-frequency queries and 0.14% for head queries. This demonstrates that semantic retrieval can significantly improve recall on hard queries without disrupting well-served popular ones. According to the authors, this is one of the largest search-quality improvements ever deployed on the platform.