Shopping by algorithm: How agentic AI deploys human heuristics as a surrogate consumer

Shopping by algorithm: How agentic AI deploys human heuristics as a surrogate consumer

Source: <https://arxiv.org/abs/2609.28372&gt;

Authors: Davood Wadi, Yu Ma

Published: 2026-09-23

Categories: econ.GN, cs.AI

arXiv: 2609.28372

Abstract

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.

Summary

Uses a Tool-Lab paradigm (product attributes behind costly tool calls) to test how just-below pricing and promotional framing influence 8 commercial LLM shopping agents. Under zero acquisition cost, pricing cues rarely mislead. Under acquisition cost, vague-goal prompts cause LLMs to skip diagnostic attributes, compute wrong unit prices, and mimic human heuristics. Specific-goal prompts preserve diagnostic search and choice optimality. Concludes that marketing heuristics in delegated AI shopping are governed by storefront information architecture rather than immutable LLM flaws.

摘要

用 Tool-Lab（产品属性藏于付费工具调用之后）检验“剑下价格”与俈惑性框察如何影响 8 个商用 LLM 购物代理。零调用成本下，价格提示几乎不会误导；加上调用成本后，模糊目标 prompt 会让 LLM 跳过诊断属性、计错单价，重现人类启发式偏误。明确目标 prompt 则保留诊断性检索与准确选择。论文认为：代理购物中的市场启发式偏误由商店信息架构决定，而非 LLM 本身的不可变缺陷。