Deep Learning Modeling of Persuasion by ChatGPT-Powered Virtual Influencers: An Artificial Neural Network MLP Test

Using an artificial neural network (ANN) multilayer perceptron (MLP) analysis, this study examined predictors of AI users’ behavioral intention to buy products recommended by ChatGPT-enabled virtual influencers. The findings from an ANN MLP analysis demonstrated that trust in ChatGPT, empathy, perceived convenience, perceived superior functions, loneliness, existential threats to human identity, education level, and socio-economic status are predictors (input layers) of AI users’ intention to buy items endorsed by ChatGPT-enabled virtual influencers (output layer). A comparative analysis between multiple linear regression (MLR) and ANN MLP methods indicates 9.88% increase in the explained variance (R2) in purchase intention and a significant reduction in the prediction error (MLR RMSE of 1.02041 to MLP RMSE of 0.3784 for the training set and 0.3689 for the testing set). The results provide strong evidence for the methodological advantages and outperformance of the deep learning ANN MLP over the MLR in the current study for modeling and capturing intricate and non-linear relationships among the variables, by demonstrating improved model fit, more variance explained, and reduced prediction error. Theoretical contributions to the burgeoning literature on AI-enabled virtual personas are highlighted. AI-powered virtual personas as business model innovation and evolving monetization frameworks are explicated in view of AI-driven digital transformation. Practical suggestions, management slant, and executive insights for AI-empowered digital marketing and synthetic personas embedded within the creator economy are also discussed.