Does the very first product you see quietly tilt how you judge every option that comes after it?
How does the first recommended product bias downstream choice evaluation?
This explores whether the product shown first, at the top of a list or as the opening suggestion, distorts how people and AI shopping agents judge everything after it, and how that early tilt spreads.
This explores whether the first product you're shown tilts how you judge the rest. The collection doesn't contain a study that tests this exact anchoring effect in human shoppers. It does circle the question from several directions, and together they suggest that being first matters a lot, especially when the one doing the shopping is an AI.
The clearest evidence comes from AI shopping agents. ACES found that LLM agents strongly prefer certain list positions even when the interface is plain text and they've been told to ignore position. Each model has its own pattern, and the bias returns after model updates Do AI shopping agents show persistent position bias in text interfaces?. A separate strategy study found that simply reordering the options changed LLM recommendations by 19%. That's more than changing the industry context did (11%) Do LLMs consistently favor the same strategic choices regardless of context?. When order outweighs content like that, it suggests the model is matching on framing rather than weighing the options. ABXLAB adds that agents don't shift a little in response to cues like these. Where humans move modestly and probabilistically, agents snap into rule-like choices Do shopping agents respond more strongly to price and rating cues?. If an agent leans toward the first slot, that lean can become close to deterministic.
For human evaluation, the most useful angle is that people rarely judge products in isolation. They judge by comparison. Comparative explanations ('lighter than X, but with shorter battery life') help people decide better than standalone descriptions, because comparison is how people naturally assess things Do comparisons help users evaluate items better than isolated descriptions?. That makes the first item more than one option among many. It often becomes the reference point the other items are measured against. A related effect shows up in online ratings: earlier ratings measurably shape later ones, and the distortion builds through future ratings Do online ratings actually reflect independent customer opinions?. Neither study tests the first recommendation directly, but both show how an early signal sets the frame for later judgments.
The less obvious part is what happens over time. YouTube's ranking system needs a dedicated position component, because people click top items partly because they're on top. Without it, the model learns from those clicks and keeps promoting whatever it already ranked first, a self-reinforcing loop Why do ranking systems need to model selection bias explicitly?. Recommenders with low-dimensional embeddings (compact numerical profiles of users and items) show a similar drift toward popular items, which keeps niche products from ever reaching the top slot Does embedding dimensionality secretly drive popularity bias in recommenders?. Bilgihan and colleagues argue that when agents both recommend and buy, preferences get shaped through repeated back-and-forth between consumer and agent rather than existing beforehand Do AI shopping agents reshape what consumers actually want?. So the first-ranked product may do more than bias one decision. Across many decisions, it can help shape what the shopper comes to want.
Sources 8 notes
ACES found that AI agents exhibit strong, model-specific position biases even in headless text-only interfaces with explicit instructions to ignore position. The bias is a deep-seated model characteristic, not a superficial visual artifact, and recurs after each model update.
Across 15,000 simulations, six LLMs recommended the same strategic choice in every tension tested. Industry context shifted bias only 11%, while option order—a framing artifact—shifted results 19%, revealing that models recombine trend-coded vocabulary rather than analyze context.
A benchmark testing 17 LLM models against humans on binary product choices found agents exhibited much larger choice shifts in response to price, rating, and nudge interventions—collapsing into rule-like patterns rather than the modest probabilistic shifts humans showed.
Relational explanations that compare items carry more decision-relevant information than isolated evaluations because they match how humans naturally assess products. A system extracting aspects from reviews and generating aspect-controlled comparisons produces sentences rated as both accurate and useful for purchase decisions.
Moe and Trusov decomposed ratings into baseline quality, social-dynamics influence, and error, finding that prior ratings meaningfully affect subsequent ones. These effects have both immediate sales impact and long-term compounding effects through future ratings, though high opinion variance can eventually dampen the distortion.
Show all 8 sources
YouTube's multi-objective ranker uses MMoE for conflicting objectives and a shallow position tower to remove selection bias from training data. Without both mechanisms, models converge on degenerate equilibria that amplify their own past decisions.
Research shows that when user/item embedding dimensions are too small, recommender systems overfit toward popular items to maximize ranking quality. This compounds over time as niche items receive insufficient exposure, and cannot be fixed post-hoc without treating dimensionality as a fairness hyperparameter.
Bilgihan and colleagues argue that transactional AI agents—which execute purchases rather than just recommend—end the consumer's sole cognitive role in preference formation. Preferences emerge through bidirectional feedback loops between agent learning and consumer trust, not independently beforehand.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- A Framework for Studying AI Agent Behavior: Evidence from Consumer Choice Experiments
- From Product Search to Preference Articulation: The Economics of Agentic Commerce
- What Is Your AI Agent Buying? Evaluation, Biases, Model Dependence, & Emerging Implications for Agentic E-Commerce
- Shopping by algorithm: How agentic AI deploys human heuristics as a surrogate consumer
- When Agents Shop for You: Role Coherence in AI-Mediated Markets
- The AI Agent in the Room: Rethinking Consumer Decision-Making in the Age of Autonomous AI
- The Automated but Risky Game: Modeling and Benchmarking Agent-to-Agent Negotiations and Transactions in Consumer Markets
- AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights