The AI wave has hit the search community hard, with semantic search becoming a hot topic of discussion. As a search engineer, you've probably engaged with it too, developing PoCs or maybe even conducting online A/B testing in production. However, it's not uncommon to conclude that ""Semantic search doesn't quite fit our domain,"" leading back to the familiar territory of lexical search, with its established synonym dictionary and knowledge graph. But what are the reasons behind the underperformance of semantic search both in offline and online? This talk delves into these critical questions, sharing insights from several years of trial and error. We will begin with a rigorous comparison of lexical and semantic search from various angles, including the validity of their evaluation metrics. A quantitative analysis will then guide us through their strengths and weaknesses, providing a balanced view of both systems. An important part of this discussion will be the integration of semantic search into existing lexical search systems, and perhaps most crucially, we will tackle an often-overlooked aspect: Is the investment in semantic search integration truly cost-effective? Attendees will gain a clearer understanding of the semantic search landscape, equipped with the knowledge to make informed decisions about its implementation in their projects.