["Psychology &Marketing, EarlyView. ", "\nABSTRACT\nDespite the proliferation of artificial intelligence (AI) in marketing, consumers often exhibit algorithm discounting—a tendency to evaluate AI‐generated recommendations less favorably than those from humans. Focusing on consumption goals, this study investigates how anthropomorphic AI agents (virtual assistants vs. robots) interact with product function (utilitarian vs. hedonic) and service target (self vs. friend) to shape consumer responses. Across three experiments combining behavioral and event‐related potential (ERP) data, AI agents consistently underperformed human recommenders across behavioral indicators—purchase likelihood, personal interest, and tipping—with each capturing distinct facets of consumer response. The neural data revealed a dissociation broadly consistent with a dual‐process interpretation: robots elicited larger N2 amplitudes (associated with early expectancy‐related conflict processing) when recommending hedonic products to friends, whereas virtual assistants did not, indicating that anthropomorphic design choices may shape early expectancy‐related responses. However, both AI types triggered heightened P3 responses (associated with controlled attentional allocation) in the same hedonic gift‐giving contexts, revealing persistent cognitive engagement regardless of agent form. This dissociation advances our understanding of when and why distinct neurocognitive processes drive consumer response to AI, and suggests that anthropomorphism may function as a double‐edged sword whose effects depend on the alignment between agent characteristics and consumption goals.\n"]