Abordagem Híbrida de Inspeção: Avaliação da Eficácia de Grandes Modelos de Linguagem no Diagnóstico de Acessibilidade Web

• Gabriel Aguiar A. e Silva UFLA
• Paulo H. dos Anojs Silveira UFLA
• Lucas Gomes Colombo UFLA
• Vanessa Fátima de Souza UFLA
• Julio Cesar Sousa de Lima UFLA
• André Pimenta Freire UFLA

DOI: https://doi.org/10.5753/webmedia.2026.30915

Resumo

Evaluating web accessibility in systems with a large volume of screens is a costly and challenging process, demanding strategies that promote greater productivity without sacrificing quality. This work investigates the use of Large Language Models (LLMs) as a tool to support manual inspection, testing their ability to generate initial estimates of problems in interfaces that share common elements. To this end, an experimental evaluation was carried out in which an LLM analyzed a set of 5 new screens based on input from previously consolidated diagnoses of 25 other screens from the same ecosystem. The results demonstrate the scenarios in which the models were able to correctly identify accessibility barriers compared to 100% manual inspection, as well as the situations that induced hallucinations or inaccuracies. The findings help to position the role of LLMs in increasing productivity within web engineering processes focused on accessibility, highlighting the indispensable care required when integrating generative artificial intelligence into the workflow of specialists.

Palavras-chave: Acessibilidade Web, Grandes Modelos de Linguagem, Inspeção de Acessibilidade

Referências

ABNT. 2024. ABNT NBR 17225: Acessibilidade em sistemas de informação e comunicação. Technical Report. Associação Brasileira de Normas Técnicas, Rio de Janeiro.

William Cain. 2024. Prompting Change: Exploring Prompt Engineering in Large Language Model AI and Its Potential to Transform Education. TechTrends 68, 1 (1 2024), 47–57. DOI: 10.1007/s11528-023-00896-0

Lúcio Cauper Freitas de França, Lisieux Marie M dos S Andrade, and Windson Viana de Carvalho. 2025. Um Estudo Sobre o Uso de Prompts LLM para a Avaliação da Acessibilidade Web dos Portais das Prefeituras da Região dos Sertões dos Crateús. In Brazilian Symposium on Multimedia and the Web (WebMedia). SBC, 580–584.

Guilherme Guerino, Luiz Rodrigues, Bruna Capeleti, Rafael Ferreira Mello, André Freire, and Luciana Zaina. 2025. Can gpt-4o evaluate usability like human experts? a comparative study on issue identification in heuristic evaluation. In IFIP Conference on Human-Computer Interaction. Springer, 381–402.

Ziyao He, Syed Fatiul Huq, and Sam Malek. 2025. Enhancing Web Accessibility: Automated Detection of Issues with Generative AI. Proceedings of the ACM on Software Engineering (2025). DOI: 10.1145/3729371

Francisco Daniel Hernández Castillo and Steffen Staab. 2025. AccessGuru: Leveraging LLMs to Detect and Correct Web Accessibility Violations in HTML Code. DOI: 10.48550/arxiv.2507.19549, arXiv preprint.

Mahmoud Khalil and Umair Rehman. 2026. Can Multimodal LLMs Model Expert Ratings in Mobile UI Usability Evaluation?. In International Conference on Human-Computer Interaction. Springer, 202–216.

Jisu Ko, Jinyoung Choi, Cielo Morales, Dajung Kim, and Minsam Ko. 2026. Criticmate: Stagewise Human–AI Co-Critique in Single-Screen UI Evaluation. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–22.

Sebastian Lubos, Alexander Felfernig, Damian Garber, Viet-Man Le, and Manuel Henrich. 2026. Recommending Usability Improvements with Multimodal Large Language Models. Proceedings of the ACM on Software Engineering 3, FSE (2026), 568–589.

Juan-Miguel López-Gil. 2024. Turning manual web accessibility success criteria into automatic: an LLM-based approach. Universal Access in the Information Society (2024). DOI: 10.1007/s10209-024-01108-z

Fabio Paternò, Manuela Vinci, and Marco Manca. 2025. How an LLM Can Improve Automatic Web Accessibility Validation?. In Proceedings of the ACM. DOI: 10.1145/3750069.3750310

Redação Forbes Brasil. 2026. Do ChatGPT ao DeepSeek: As 10 IAs mais Utilizadas por Profissionais. [link]

Pranab Sahoo, Ayush Kumar Singh, Sriparna Saha, Vinija Jain, Samrat Mondal, and Aman Chadha. 2024. A Systematic Survey of Prompt Engineering in Large Language Models: Techniques and Applications. arXiv:2402.07927 [cs.AI]

Shristi Shrestha and Anas Mahmoud. 2026. Leveraging GPT-4o to Evaluate the Usability and Accessibility of Flutter App Interfaces. ACM Transactions on Software Engineering and Methodology (2026).

Radu-Daniel Vatavu. 2025. When LLM-Generated Code Perpetuates User Interface Accessibility Barriers, How Can We Break the Cycle?. In Proceedings of the ACM. DOI: 10.1145/3744257.3744266

Guillermo Vera-Amaro and Jose Rafael Rojano Caceres. 2025. Accessible Web Content Generation Using LLMs: An Empirical Study on Prompting Strategies and Template-Guided Remediation. IEEE Latin America Transactions (2025). DOI: 10.1109/tla.2025.11231224

Markel Vigo, Justin Brown, and Vivienne Conway. 2013. Benchmarking web accessibility evaluation tools: measuring the harm of sole reliance on automated tests. In Proceedings of the 10th international cross-disciplinary conference on web accessibility. 1–10.

W3C. 2023. Web Content Accessibility Guidelines (WCAG) 2.2. Technical Report. World Wide Web Consortium. [link]

Jules White, Quchen Fu, Sam Hays, Michael Sandborn, Carlos Olea, Henry Gilbert, Ashraf Elnashar, Jesse Spencer-Smith, and Douglas C. Schmidt. 2023. A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT. DOI: 10.48550/arXiv.2302.11382 arXiv preprint.

Lucas Emanuel Oliveira Xavier, Daniel Mesquita Feijó Rabelo, Ribamar de Souza Martins, and Windson Viana. 2025. Image-to-UI Accessibility: Assessing ChatGPT’s Effectiveness in Generating Accessible Android Screens from Figma Templates. In Brazilian Symposium on Multimedia and the Web (WebMedia). SBC, 302–311.