Answers are grounded in this paper and cite their sources.
AI-generated review summaries may omit infrequent reports relevant to consumer safety judgements. Guided by the accessibility–diagnosticity framework, three studies examine information selection and consumer responses. Study 1 links 49,009 pre-recall Amazon reviews from 53 product–recall groups to CPSC records and generates 9680 summaries from 2420 review windows across four language models. Study 2 compares 228 summaries for 38 recalled products under common-theme, general-diversity and risk-priority rules. Study 3 uses a 2-by-2 online experiment (N = 330) to examine summary mention and issue type while keeping the target source review visible. More frequent safety signals were more likely to enter summaries under recurring-experience instructions. A risk-priority rule retained isolated safety reports more often than a common-theme rule. In Study 3, summary mention increased the corresponding risk perception: physical safety risk for safety issues and performance risk for ordinary quality issues. Purchase intention decreased significantly in the ordinary quality condition, but not in the safety condition. Estimated indirect paths through the corresponding risk dimensions provided evidence consistent with the proposed mechanism. The findings distinguish the retention of specific safety reports from overall review representativeness and show that summary presentation can affect issue-specific risk judgements even when the source report remains available.
No takes yet. Share an insight, caveat, or question.
Guo et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: