Marketing Keywords Exploration via LLMs With Domain Expertise
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
dc.identifier.issn
2169-3536
dc.identifier.doi
10.1109/ACCESS.2026.3659535
dc.identifier.other
dc.description.volume
14
dc.description.issue
Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI
10.1109/ACCESS.2026.3659535
10.1109/ACCESS.2026.3659535
10.1109/ACCESS.2026.3659535
Abstract
We demonstrated how to explore the space of search marketing keywords using LLMs with the help of tool support. These external tools represent the domain expertise and the use of proprietary performance data. The LLM consults the tools provided when necessary in order to generate keywords that are expected to get user attention. We investigated whether the linguistic expressiveness of an LLM when combined with valid purchase intents provided by human marketers as few shots in a prompt would result in discovering new prospective keywords. We conducted an empirical study on two different domains: 1) the healthcare domain and 2) the online learning domain. The LLM agent was able to explore a viable set of keywords, which was validated quantitatively by the number of user impressions, clicks and conversions the generated keywords were expected to receive.
Description
Keywords
Retrieval augmented generation , Vectors , Search engines , Medical services , Transformers , Training , Space exploration , Engines , Data models , Collaboration , Agentic artificial intelligence , large language models , search engine marketing , Computer Science, Information Systems , Engineering, Electrical & Electronic , Telecommunications , Computer Science , Engineering