Choosing JavaScript packages requires developers to search large software repositories, compare alternatives, and rank candidates according to technical needs. This work studies whether a large-language-model-based assistant can support that task and compares it with AIDT, a recommender that uses meta-search and machine learning to identify relevant packages. The comparative experiment examines the quality of package recommendations and how well the approaches align with developers’ expectations. The results provide evidence about where LLM-based assistance can complement conventional recommender techniques and where more specialized retrieval and ranking mechanisms remain advantageous for technology selection.