When an AI agent chooses for you: price, ratings and nudges move its decisions too
Agents have no human cognitive limits, yet they are still biased choosers — change how a price is framed, where a rating comes from or how a nudge is worded, and the choice changes.
Key numbers
- Framework: ABxLab probes agentic choice in a realistic web shopping environment by systematically varying option attributes (price, ratings) and persuasive cues (psychological nudges).
- Agent decisions shifted predictably and substantially in response — biased choosers even without the cognitive constraints that shape human biases.
- The authors frame it as both risk and opportunity: risk because agentic consumers may inherit and amplify human biases; opportunity because consumer choice is a powerful testbed for a behavioural science of AI agents.
- The framework is released as an open benchmark for scalable evaluation of agent decision-making.
Method and sample
arXiv paper “A Framework for Studying AI Agent Behavior: Evidence from Consumer Choice Experiments” (2025-09-30): the ABxLab framework varies price, ratings and persuasive cues in a web shopping environment to measure shifts in LLM agent choices, released as an open benchmark.
What it means for brands
If agents increasingly make purchase decisions, being recommended stops being purely about ranking: how price is presented, where ratings come from and how wording nudges all influence the agent. Brands need to understand where they sit in that agent-readable competition.
How SourceSight handles it
SourceSight records what AI actually answers and cites, so differences in how the same prompts describe your brand across engines become visible — the first step in checking whether your claims survive AI mediation.
Limits
The work runs in controlled shopping environments rather than live transactions; ABxLab observes behavioural shifts and does not model real payment, inventory or supply constraints.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
What 2 million AI citations show: prompt-content match matters most, checklists matter less
Once you control for the domain, the usual AEO checklist (FAQ blocks, structured data, Core Web Vitals) reverses or collapses to zero — domain authority and prompt alignment dominate.
How engines work · 2026-10-07Whose side is the AI shopping assistant on? Change its role and the penalty for sponsored results disappears
The same sponsored listings get treated very differently once the prompt names the platform as the agent's principal — disclosure rules built for humans do not carry over.