Apple researchers conducted an A/B test using AI to refine App Store search rankings, achieving a +0.24% lift in conversion rates. The study, published in *Scaling Search Relevance*, focused on improving how apps surface in search results by combining behavioral data (user interactions) with AI-generated textual relevance labels. While human-curated textual labels are scarce, Apple’s 3-billion-parameter LLM filled the gap, generating millions of labels to retrain its ranking algorithm.

Results showed the LLM-augmented model outperformed traditional methods in 89% of global storefronts, translating to tens of millions of additional app downloads annually if scaled.