From Compliance to Decision Confidence: A Scoping Review of Accessible E-Commerce for Blind and Low-Vision People
Bektur Ryskeldiev, Matthew Gillingham, Norimasa Kobori · 2026 · Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA '26) · doi:10.1145/3772363.3799042
Summary
Ryskeldiev, Gillingham, and Kobori (Mercari R4D and University of Tsukuba) present a scoping review of 21 papers (14 directly about e-commerce accessibility for blind and low-vision (BLV) users, 7 adjacent) published 2010-2025, drawn from ACM DL, Google Scholar, and Springer/Elsevier. The central reframing argument is that accessibility research and practice both overweight WCAG compliance — code-level violations — while the lived problem for BLV shoppers and sellers is decision confidence: can a user judge product suitability and transaction risk well enough to buy or sell without ad hoc sighted assistance? Recent industry data support the critique: a 2025 Contentsquare audit found 84% of checkout pages on leading e-commerce sites are functionally invisible to screen readers, and WebAIM's 2025 Million report shows 95% of home pages fail WCAG. The authors map the BLV e-commerce journey into five stages — Discover, Evaluate, Transact, Fulfill, Sell — and observe that 10/14 direct papers address buyer tasks while only 1/14 proposes a seller-focused technical intervention. They derive four challenge clusters (visual content, navigation, fulfilment, trust) and cross them against five research approaches (GenAI/LLMs, computer vision, wearable/telepresence, audits/guidelines, qualitative studies) in a challenge-approach matrix that makes research gaps visible at a glance.
Key findings
The review identifies three major gaps. First, seller-side workflows are almost entirely unsupported: one BLV seller reported spending three hours listing a single item and outsourcing product-photo verification through Be My Eyes — a decision-confidence failure that cascades from buyer tools into seller burden, limiting entrepreneurship and gig-economy participation. Second, generative AI has been applied to buyer tasks (one paper) but not to seller tasks such as product photography guidance, listing generation from seller narration, or inventory description. Third, trust and verification rely on user-coordinated strategies (cross-checking with Be My AI, remote sighted assistants, platform reviews) rather than systematised tools — no paper in the corpus proposes automated verification infrastructure, which is precisely what BLV users resort to ad hoc. The challenge-approach matrix also shows empty cells for wearable/telepresence work on visual-content tasks and for fulfilment-stage GenAI, both of which the authors flag as productive directions. The paper's headline reframing is that WCAG conformance measures code violations while decision confidence measures whether transactions can actually be completed — the two only loosely correlate, and compliance-focused metrics can hide real accessibility failures.
Relevance
This review offers three things of direct practical value to accessibility practitioners. First, it gives a vocabulary (decision confidence, challenge-approach matrix, buyer-journey stages) that can be used to justify accessibility spend beyond compliance — useful when arguing with legal and procurement teams who treat WCAG as the ceiling. Second, the identification of seller-side accessibility as a near-total gap is actionable: C2C platforms (eBay, Etsy, Mercari, Facebook Marketplace, Vinted, Depop), gig-work platforms (Uber, DoorDash, TaskRabbit), and small-business SaaS (Shopify, Squarespace) should invest in accessible listing flows, photo-guidance AI, and inventory management tooling if they do not want to continue excluding disabled sellers. Third, with the European Accessibility Act applying to e-commerce services from June 2025, the paper's argument that compliance does not equal usability is already regulatorily relevant: organisations that pass a technical audit but fail BLV users at decision-critical steps remain exposed. Limitations: the corpus is small, English-language only, from HCI venues (so misses payments, logistics, and legal literatures), and coded by a single researcher. Commercial audits of Amazon, Shopify, and eBay are absent from the synthesis. The authors position the contribution as a mapping of themes and gaps rather than a complete assessment.
Tags: e-commerce accessibility · blind and low vision · scoping review · online shopping · C2C marketplaces · generative AI · screen readers · European Accessibility Act · accessibility research · gig economy
Standards referenced: WCAG 2.2 · European Accessibility Act