How Yahoo enhances search retargeting using Amazon Bedrock
In this post, we demonstrate how Yahoo implemented Amazon Bedrock to enhance their Search Retargeting (SRT) capabilities in the Yahoo DSP ad tech suite. SRT is a core audience targeting solution that helps advertisers reach users based on their historical search behavior, bridging search intent with display, video, and native advertising. Beyond targeting keywords entered on Yahoo Search, SRT uses AI to identify and engage users who demonstrate intent through search activity both on Yahoo and across integrated partner systems.
Follow AWS Machine Learning Blog to make it a durable For You signal.
Yahoo replaced its legacy keyword expansion system for Search Retargeting (SRT) with a generative AI pipeline built on Amazon Bedrock, specifically using Anthropic’s Claude 3.5 Sonnet v2. The previous method relied on Word2Vec embeddings and locality-sensitive hashing, which often produced zero expansions, relied on syntactic rather than semantic similarity, and used outdated vocabulary. The new LLM-based approach generates semantically relevant keyword sets and includes a guardrailing step: each expanded keyword is embedded and checked against the original using cosine similarity, with a quality filter and sensitive-term blocks applied both before and after model inference. In production since Q1 2025, the system yields up to a 600x increase in keyword expansion rates, up to 5x growth in addressable audience reach, a fivefold improvement in median broad expansion ratio, and a doubling of the maximum expansion ratio under strict similarity thresholds. This matters because it lets advertisers reach larger, more intent-aligned audiences across display, video, and native channels while maintaining policy compliance and avoiding hallucinations, strengthening Yahoo DSP’s competitive ad–