New AI Platform Generates University Course Content Aligned with Learning Standards

The platform, called iLeapTutor, is described in a paper published in Frontiers in Computer Science as a proof-of-concept tested on a single course, Selected Topics in Computer Science, at the university.

The system is designed to address a familiar problem in digital education: producing structured, course-aligned learning materials that satisfy formal requirements such as Course Learning Outcomes (CLOs) and the standards of Saudi Arabia’s National E-learning Center (NeLC). Rather than leaving an AI model to generate content freely, iLeapTutor anchors its output to the tutor’s own uploaded course materials, so that the generated lessons remain consistent with what the human instructor has approved.

In this case, the underlying model is Gemini 3.1 Flash-Lite (Preview). The prompts sent to the model embed the CLO and NeLC requirements directly, and the tutor then reviews and approves the generated course before it is used, keeping a human in the loop for quality control.

The evaluation combined automated metrics with expert judgment. On the automated side, the generated content achieved a BERTScore-F1 of 0.829, a measure of semantic similarity between the generated text and the reference material. Consistency was assessed with Self-BLEU: a cross-run score of 0.553 and a within-lesson score of 0.146, indicating that repeated generations of the same course stayed similar to one another while individual lessons retained internal diversity.

Latency was also measured. Generating the full set of 16 course topics took an average of 219 seconds, a figure that reflects the scale of the task rather than the speed of a single response.

Thirteen academic experts then evaluated the generated course across five dimensions: content accuracy, alignment with the Course Learning Outcomes, structure, language quality, and platform usability. At least 92.3 percent of responses on every questionnaire item indicated agreement, and all thirteen experts said they would recommend the platform.

The study’s authors are careful to frame the work as a single-course proof of concept rather than a general-purpose solution. The evaluation was conducted on one course at one institution, and the paper does not claim that the same results would hold across other disciplines, languages, or institutional standards.

The combination of retrieval grounding, embedded regulatory requirements, and mandatory human review is presented as the mechanism that keeps the generated material aligned with both the course and the national e-learning framework.

The research was received on 10 September 2026 and accepted on 12 October 2026, and the final formatted version of the article is scheduled for publication in the journal’s Digital Education section. The work was carried out by a team from the Computer Science Department at Imam Abdulrahman Bin Faisal University in Jubail, with one co-author from Shaqra University.

As generative AI tools continue to move into education, studies like this one offer a measured data point: automated metrics suggest strong semantic alignment and output consistency, expert reviewers responded positively, and the platform’s latency is reported in seconds rather than hours. Whether those results generalize beyond a single computer science course remains an open question that the authors’ own framing leaves for future work.

Subject of Research: A generative AI platform for creating standards-aligned university course materials

Article Title: iLeapTutor: a RAG-enabled generative AI platform for standards-aligned university course authoring

Article References: Alboaneen, D., Alanzi, N., Alrshidi, N., Aljearah, W., Alshahrani, S. M., Alsayednasser, F., & Alibrahim, L. (2026). iLeapTutor: a RAG-enabled generative AI platform for standards-aligned university course authoring. Frontiers in Computer Science, 8, 1994444. https://doi.org/10.3389/fcomp.2026.1994444

Image Credits: AI Generated

DOI: 10.3389/fcomp.2026.1994444

Keywords: generative AI, intelligent tutoring systems, retrieval-augmented generation, course authoring, learning outcomes alignment, higher education, large language models, digital education, BERTScore, Saudi Arabia, iLeapTutor, RAG-enabled