This paper explores the performance of three state-of-the-art large language models (LLMs)—GPT-3.5, GPT-4, and Llama 2—in sentiment analysis tasks compared to traditional transfer learning models. The study investigates how textual data characteristics and analytical procedures influence classification accuracy, and evaluates the explainability of sentiment classifications generated by LLMs. The findings reveal that LLMs can compete with and in some cases surpass traditional transfer learning methods in sentiment classification accuracy. The performance of LLMs is influenced by factors such as the presence of lengthy words, the origin of the data, and the complexity of the text. Llama 2 is found to provide the best classification explanations, demonstrating advanced human-like reasoning capabilities. The study also highlights the importance of data characteristics and the impact of prompting methods on LLM performance. The results suggest that while LLMs offer significant potential in sentiment analysis, specialized transfer learning models like RoBERTa can outperform them in certain contexts. The study provides valuable insights for marketing researchers and practitioners in selecting appropriate methods for sentiment analysis in the age of Generative AI.This paper explores the performance of three state-of-the-art large language models (LLMs)—GPT-3.5, GPT-4, and Llama 2—in sentiment analysis tasks compared to traditional transfer learning models. The study investigates how textual data characteristics and analytical procedures influence classification accuracy, and evaluates the explainability of sentiment classifications generated by LLMs. The findings reveal that LLMs can compete with and in some cases surpass traditional transfer learning methods in sentiment classification accuracy. The performance of LLMs is influenced by factors such as the presence of lengthy words, the origin of the data, and the complexity of the text. Llama 2 is found to provide the best classification explanations, demonstrating advanced human-like reasoning capabilities. The study also highlights the importance of data characteristics and the impact of prompting methods on LLM performance. The results suggest that while LLMs offer significant potential in sentiment analysis, specialized transfer learning models like RoBERTa can outperform them in certain contexts. The study provides valuable insights for marketing researchers and practitioners in selecting appropriate methods for sentiment analysis in the age of Generative AI.