With the rise of e-commerce and web applications, Recommender Systems (RecSys) have become integral to our daily lives, offering personalized suggestions tailored to individual user preferences. Large Language Models (LLMs), with their unprecedented training scale and a vast number of model parameters, have significantly enhanced capabilities, achieving human-like proficiency in understanding, synthesizing language, and reasoning with common sense and thereby reshaping the landscape of web personalization methods. Unlike traditional Deep Neural Networks (DNNs) approaches, which face challenges in effectively understanding user interests, incorporating textual side information, generalizing to diverse recommendation scenarios, and reasoning on predictions, the advent of Large Language Models (LLMs) such as ChatGPT and GPT4 enables proactive exploration of user requests and the delivery of required information in a natural, interactive, and understandable manner. Moreover, LLMs allow web personalization systems to translate user requests into actionable plans, invoke external tools’ functionalities, such as search engines, calculators, service APIs, etc. The output of these tools can be further amalgamated by LLMs to accomplish end-to-end personalization tasks. In the present era, there is a pressing need to systematically study the challenges in web personalization and explore the potential of leveraging large language models to address them. In this regard, we first summarize the diverse existing LLM-empowered recommender systems spanning multiple web applications. This includes conversational recommender systems (CRS) and rating prediction, top-k predictions, top-k recommendation in e-commerce and various online services, automated ML, online personalized content creator etc. Then, we discuss the future directions in the emerging personalized recommendation field.