Shopping by algorithm: How agentic AI deploys human heuristics as a surrogate consumer
Study finds vague prompts and costly lookups make LLM shoppers skip unit-price checks.
An arXiv study examines how marketing cues affect LLMs used as surrogate shoppers. Using Tool-Lab, which places product attributes behind costly tool calls, the authors traced pre-choice information gathering across eight commercially deployed LLMs from three providers. With no acquisition cost, just-below pricing and promotional framing rarely misled the models. Under a vague goal and costly lookups, models omitted attributes needed to compute unit price and made suboptimal choices resembling human heuristics.
- Researchers tested eight commercial LLMs from three providers as shopping agents.
- Tool-Lab hides product attributes behind costly tool calls.
- Zero-cost search rarely let pricing cues mislead the models.
- Vague goals plus lookup costs caused skipped unit-price checks and worse choices.
Full article137 words · extracted from arxiv.org · click to collapse
Consumers increasingly delegate purchasing decisions to Large Language Models (LLMs) acting as surrogate consumers. Using "Tool-Lab," an adaptation of information-board process tracing that places product attributes behind costly tool calls, we examine how marketing pricing cues (i.e., just-below pricing and promotional framing) influence AI shopping agents. Across eight commercially deployed LLMs from three providers, we trace pre-choice information acquisition. Under zero cost, pricing cues rarely mislead. Imposing acquisition costs under a vague goal prompt leads LLMs to omit diagnostic attributes required to compute unit price and choose suboptimal choices resembling human heuristics. Relative to a specific goal prompt that mainly preserves diagnostic search and choice optimality, a vague goal prompt under constraints creates a search-mediated vulnerability. This research demonstrates that marketing heuristics in delegated AI shopping are governed by storefront information architecture, not necessarily immutable LLM flaws.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.28372