Generative artificial intelligence chatbots in investment decision-making: a phantom menace or a new hope?

What the paper says

Purpose This study aims to investigate the relevance, accuracy, specificity and justification of investment recommendations of generative artificial intelligence (GenAI) chatbots for different investment capitals and countries (UK and Bulgaria). Design/methodology/approach A two-stage mixed methods approach was used. Prompts were queried into OpenAI’s ChatGPT, Microsoft Bing and Google Bard (now Gemini). Finance and investment practitioners and finance and investment lecturers assessed the chatbots’ recommendations through an online questionnaire using a five-point Likert scale. The Chi-squared test, Wilcoxon-signed ranks test, Mann–Whitney U test and Friedman test were used for data analysis to compare GenAIs’ recommendations for the UK and Bulgaria across different amounts of investment capital and to assess the consistency of the chatbots. Findings GenAI chatbots’ responses were found to perform medium-to-high in terms of relevance, accuracy, specificity and justification. For the UK sample, the amount of investment had a marginal effect but prompt timing had an interesting impact. Unlike the British sample, the GenAI application, prompt timing and investment amount did not significantly influence the Bulgarian respondents’ evaluations. While the mean responses of the British sample were slightly higher, these differences were not statistically significant, indicating that ChatGPT, Bing and Bard performed similarly in both the UK and Bulgaria. Originality/value The study assesses the relevance, accuracy, specificity and justification of GenAI chatbots’ investment recommendations for two different periods, investment amounts and countries.