Academic library online chat services under the impact of artificial intelligence

Purpose This study aims to explore the capabilities, limitations and potential of ChatGPT applicable to online reference services in academic libraries. Design/methodology/approach This study used the method of qualitative content analytics to assess the general capabilities of ChatGPT applicable in academic libraries. Two experienced academic librarians had face-to-face interactions with ChatGPT by asking ten most common questions often asked by faculty and students at the Georgia Southern University Libraries (https://library.georgiasouthern.edu/). To examine the ChatGPT’s applicability and capability, they also compared the ChatGPT with a popular online chat reference tool called LibChat, which is now widely used in academic libraries in 91 countries worldwide. Findings It was found that as an artificial intelligence (AI)-powered real-time chatbot ChatGPT could effectively provide faculty and students with general guidance on locating the needed information resources and services in academic libraries, though its responses might not be accurate or truthful all the time. Embedded into the LibAnswers system of the Springshare’s products (www.springshare.com/libanswers/), LibChat serves as a real-time online chat tool used by academic libraries for reference services, but it is only available during the regular librarians' duty hours. This technical limitation does not meet the dynamic needs of faculty, students, staff, and local community users. Only well-optimized AI-driven chat products like ChatGPT could provide 24/7 online services to support uninterrupted academic library services in the future. Research limitations/implications This study only examined the general capability and potential of ChatGPT3.5 in specific subject areas. Additional studies are needed to further explore how the latest capabilities of ChatGPT4.0 or newer version, such as its text-to-image, text-to-speech, text-to-text, text-to-video and Web search, could impact future reference services of academic libraries. ChatGPT’s primary optimization and upgrades in the future may also change and impact this study's findings. The comparison between ChatGPT and LibChat presents a significant breakthrough of the generative AI technology in academic libraries. This comparative study encourages more academic experts, faculty, librarians and scholars to track the advance of generative AI applications, including ChatGPT, adopted in academic learning environments. In addition, the ChatGPT's complete capability and potential enhanced and integrated in the future may go beyond what this study evaluated. Originality/value This study examined the strengths and weaknesses of ChatGPT applicable to reference services of academic libraries. Through a comparison between ChatGPT and LibChat, this study suggests that optimized AI online chatbots still have a long way to go to meet the dynamic needs of faculty and students in the ever-changing academic learning environments. To contribute to the existing research literature focusing on the rise of generative AI tools such as ChatGPT, this study provides a valuable reference for the applicability of generative AI applications in academic libraries to promote more library creation and innovation in the coming years of the 21st century.