A platform to check website compliance with web accessibility standards
Reinaldo Ferraz, Ana Duarte, João Bárbara, Adriano C.M. Pereira, Wagner Meira · 2023 · Proceedings of the 20th International Web for All Conference (W4A) · doi:10.1145/3587281.3587289
Summary
This short paper presents the TIC Web Acessibilidade platform, a comprehensive system for evaluating and monitoring the accessibility of Brazilian government websites at scale. Developed by NIC.br (the Brazilian Internet Steering Committee's executive arm) in collaboration with the Federal University of Minas Gerais, the platform addresses a persistent gap in accessibility oversight for the Brazilian public web. Brazil's digital accessibility model, eMAG (Modelo de Acessibilidade em Governo Eletrônico), provides mandatory guidelines for government websites based on WCAG 2.0, covering all Level A, AA, and some AAA success criteria without conformance tiers — all recommendations are mandatory. The platform's architecture consists of three modules: a Web Collector that crawls government websites using the Scrapy library, an Asynchronous Evaluator that processes collected pages through the ASES (Avaliador e Simulador de Acessibilidade em Sítios) analysis tool, and an Accessibility Portal built on Ruby on Rails with MongoDB that aggregates and displays results. The ASES tool generates weighted accessibility scores from 0 to 100, where different error types carry different weights (HTML markup errors weight 3, CSS errors weight 1). The platform supports analysis at multiple levels of abstraction — individual pages, whole websites, and entire domains — and provides features including compliance ranges, most common errors, top-performing websites, and rating history over time.
Key findings
Applied to the Brazilian gov.br domain, the platform evaluated 1,445 websites comprising 415,117 pages. Results reveal widespread accessibility deficiencies: at launch only 2,407 pages scored above 95 (out of 267,090), and only one page achieved a perfect score of 100. By the time of writing, just seven websites had reached scores above 95. The best state-level average was only 82.41, falling in the "considerable number of errors" tier. The five most common errors were: unclear link descriptions (6.7 million errors across 398,463 pages), keyboard inaccessibility (5.6 million errors on 180,685 pages), poor HTML semantic structure (3.1 million errors on 413,001 pages), missing image alt text (2.3 million errors on 405,687 pages), and incorrect heading hierarchy (748,000 errors on 403,602 pages). The keyboard accessibility finding is notable — fewer pages were affected but with higher error density per page, likely from mouse-only event handlers without keyboard equivalents.
Relevance
This platform represents an important model for government-level accessibility monitoring, similar to Portugal's Accessibility Observatory and Italy's Vamolà project. For accessibility practitioners, the findings confirm that the most common barriers are fundamental issues — link descriptions, keyboard access, semantic HTML, alt text, and heading structure — rather than exotic edge cases. The weighted scoring approach offers a more nuanced view than binary pass/fail compliance checking, though the authors note that even a score of 95+ does not guarantee full accessibility without human verification. The platform's open-source availability (CC BY-SA 4.0) makes it adaptable for other countries seeking to monitor government web accessibility. For the Global South accessibility community, this work demonstrates both the feasibility of large-scale automated monitoring and the scale of the challenge that remains.
Tags: automated testing · web accessibility · government accessibility · accessibility monitoring · web crawler · compliance · Global South accessibility
Standards referenced: WCAG 2.0 · eMAG