Abstract:

Learning Management Systems (LMS), particularly Moodle, have become an essential infrastructure in modern education, serving millions of educators worldwide. As institutions increasingly rely on these platforms for content delivery and assessment, Artificial Intelligence (AI) integration presents transformative opportunities for enhancing pedagogical efficiency. The rapid evolution of AI technologies demands continuous reassessment of their capabilities, particularly as these tools automate content creation tasks that traditionally required technical expertise.

This follow-up study examines the performance evolution of AI platforms in automating HTML-based self-assessment form generation for Moodle, comparing results obtained within one year of an initial investigation. The research addresses creating interactive learning resources without extensive programming knowledge, thereby democratizing access to educational tools.

Building upon the original comparative study of four AI platforms (ChatGPT, Gemini, Claude, and Copilot), this research expands the analysis to eight models, incorporating DeepSeek 3.2, Grok 4.1, Qwen 3, and Mistral Large 3. The methodology maintained identical conditions—same standardized prompt, reference image, and evaluation criteria (Layout Accuracy, Calculation Functionality, Prompt Adherence, and Functionality in Moodle)—to ensure direct comparability and isolate performance changes attributable to model evolution.

The results reveal significant shifts in platform capabilities and competitive positioning. Gemini demonstrated remarkable evolution, achieving excellence across all evaluation criteria and emerging as the top-performing platform—a dramatic improvement from its previous partial functionality. Claude, while maintaining strong performance, no longer holds its previous leadership position. ChatGPT and Copilot showed substantial improvements, with ChatGPT resolving critical functionality issues and Copilot advancing from the lowest performer to the second-tier group. Among new platforms, Grok, Qwen, and Mistral achieved functional results, while DeepSeek notably underperformed with significant structural and computational failures.

These findings underscore critical insights for educational technology adoption:
(1) AI model performance evolves heterogeneously and rapidly, with some platforms advancing substantially while others stagnate within short timeframes;
(2) leadership in AI capabilities is dynamic and subject to frequent shifts;
(3) periodic reassessment of AI tools is essential to avoid outdated assumptions that could lead to suboptimal pedagogical choices;
(4) expanding comparative studies to include emerging and open-source models provides valuable insights for evidence-based tool selection.

The study reinforces that developing prompt engineering skills remains crucial for educators seeking to optimize AI-assisted content creation, while the increasing reliability of low-code AI solutions offers promising pathways for non-technical educators to create sophisticated interactive learning resources directly within Moodle environments.

Keywords:

Artificial Intelligence (AI), Learning Management Systems (LMS), MOODLE, Prompt Engineering, Model Evolution, Educational Technology.