Empowering Web Editors with Generative AI : Creating a Tool for Efficient Search Engine Optimization within a Content Management System

As content volume on the web increases, web editors face growing pressure to create high-quality, search engine-optimized content efficiently. Despite the importance of search engine optimization (SEO) for website visibility and user engagement, many websites still suffer from missing or poorly crafted metadata. This thesis explores how Generative Artificial Intelligence (Gen AI), specifically Large Language Models (LLMs), can be integrated into a Content Management System (CMS) to support Swedish web editors in the creation of SEO metadata. The project followed a user-centered Design Thinking process, involving six webeditors through interviews and testing, enriched by expert evaluations. The approach was informed by prior research on AI, SEO, and User Experience (UX), as well as a design framework for Gen AI interfaces and benchmarks on LLM performance. Based on these insights, both a high-fidelity Figma prototype and a functional CMS extension were developed and evaluated. Beyond standard interface design, the prototyping process included custom evaluations of model performance and prompt engineering to address user needs and technical constraints. The final solution featured an AI assistant, a search result preview, and a dashboard displaying performance across pages, designed to increase transparency, control, and motivation for adoption. Results show that the AI assistant achieved high usability and user satisfaction, scoring a System Usability Scale (SUS) score of 100. Although significant time savings could not be confirmed in this limited test, web editors reported perceived improvements in both efficiency and content quality. This study presents a practical methodology for designing and evaluating AI-powered metadata tools that balance the needs of a target group with diverse levels of SEO and AI expertise. Finally, it demonstrates that prototyping with GenAI is a cross-disciplinary task that requires both design and engineering, including prompt development and model evaluation, to ensure solutions are not only functional but also understood, trusted, and usable.