Lost in summarization

: how AI-generated review summaries shape consumer trust

Abstract

Artificial intelligence (AI) is transforming e-commerce, including the way consumers interact with online reviews. Major platforms, such as Amazon, now use AI-generated review summaries (AGRS) to reduce information overload. However, the effect of AGRS on consumer trust remains unclear. This thesis examines whether AGRS enhance or diminish trust, how this effect depends on review valence (positive, negative, or two-sided), and whether perceived helpfulness mediates these relationships. A 2×3 online experiment with 291 participants tested how consumers responded to reviews that differed in valence and whether or not they included an AGRS. The findings showed that the AGRS alone did not significantly influence trust or perceived helpfulness. However, helpfulness strongly predicted trust. Review valence had a partial effect: positive summaries were perceived as being less trustworthy than negative or two-sided ones. These results extend trust transfer and information processing theories by showing that AGRS have little influence in high-trust contexts, except when summarizing positive reviews. Managers should adopt AGRS only when they clearly add value, ensure transparency, and position them as complements rather than substitutes. For researchers, the results underscore the need to test AGRS in more naturalistic and demanding settings, as well as across product types, to better understand their role in consumer decision-making.

Date of Award	15 Oct 2025
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Paulo Romeiro (Supervisor)