AI Built Right into Moodle Core: What the LMS Brings to the Table in Terms of Artificial Intelligence.
Since Moodle 4.5, a complete AI subsystem has been built into the Moodle core—not a plugin or an add-on, but part of the standard installation. Four specific functions are available: text generation, image generation, content summarization, and content explanation. Versions 5.0 and 5.1 added additional providers, granular control options, and the explanation function. This article demonstrates what the system can do, how it is architecturally structured, and what control institutions retain over its use.
The AI Subsystem: The Architecture Behind the Scenes
The AI subsystem is not tied to a single AI provider. Instead, it is based on a three-tier architecture that offers maximum flexibility:
The AI Provider feature establishes a connection to external AI services. Moodle currently supports the OpenAI API and Azure AI in its core. Since the OpenAI API has become the de facto standard, self-hosted solutions such as Ollama, LiteLLM, and LocalAI, as well as cloud services like Groq, also work through it. Starting with Moodle 5.1, DeepSeek is also supported as a provider. Institutions can therefore freely choose whether to use commercial services, run open-source models on their own infrastructure, or combine both.
Action settings define what the AI can do. Each action is its own class in the system, for example, generate_text, generate_image or summarise_text. Providers specify which actions they support, and the system automatically coordinates the assignment, including fallback logic. For example, a low-cost model can be used for simple summaries, while a more powerful model is used for text generation.
The AI placement determines where AI features are visible to users in Moodle. Currently, there are two core placements: the text editor and course support.
So anyone using Moodle already has access to the AI subsystem, but still needs to select and integrate an AI provider.
The four AI features in detail
1. Generate text
An AI icon (the familiar “Sparkle” icon) appears in the TinyMCE editor, allowing trainers—and, if authorized, participants as well—to generate text. To do this, you simply enter a prompt describing the text you need, and you’ll receive an AI-generated suggestion that can be inserted directly into the editor.
Typical use cases:
Creating introductions for course pages or lessons, formulating assignments, generating discussion questions, or drafting explanatory texts on complex topics.
2. Generate images
You can also use the text editor to generate images via text prompts, for example using models like DALL·E 3. You can choose the quality (Standard/High) and format (Square, Landscape, Portrait). The generated image is automatically assigned alt text, which improves accessibility.
Typical use cases:
Creating course thumbnails and illustrations, generating infographics for educational materials, or producing visual representations of abstract concepts.
3. Summarize the content
Using the "Course Support" feature, learners can click a "Summarize" button on any course page—whether it's a text page, quiz, forum, or lesson. The AI condenses the page content into a concise overview. This is particularly helpful for long texts or complex technical vocabulary.
4. Explain the content
Since Moodle 5.0, the "Explain" feature has been available alongside the summary. It provides a clearer, alternative way of phrasing the page content, making it ideal for learners with varying levels of prior knowledge, non-native speakers, or when dealing with particularly challenging material.
Full control: Permissions at every level
Moodle takes a consistently granular approach to AI control:
- Admins enable providers and placements globally, set rate limits, and define system prompts that precede every AI request. By default, everything is disabled—AI must be explicitly enabled.
- Role-based permissions determine which user groups are allowed to perform which actions. For example, you can allow instructors to generate images and text, while participants are limited to summarizing and explaining.
- Starting with Moodle 5.1, instructors can enable or disable AI tools for each course—and even for each activity—directly in the course or activity settings, without having to navigate through complex permissions pages.
AI Usage Policy: Transparency as a Requirement
Before users use an AI tool for the first time, they are shown an AI Usage Policy, which they must actively accept. The Moodle developer documentation describes the requirement as follows: Consent must be given at the “point of use”—that is, exactly where the AI is actually being used—and not hidden within a general site policy. For this reason, a separate policy function, distinct from the regular site policy, has been implemented for the AI subsystem.
The Terms of Use only need to be accepted once: Once a user has confirmed them in one context (e.g., in the text editor), they will not be asked again in other contexts (e.g., course assistance). If a user declines the policy, the AI features will remain disabled for that user.
The policy text can be customized using the Moodle language pack. This means that institutions can adapt the text to their own AI policies.
Starting with Moodle 5.0, administrators can also view who has accepted the policy and how the AI tools are being used—including token usage, context, and user ID—via Site Administration → Reports → AI Reports. This implementation aligns with Moodle’s AI Principles, which define transparency and configurability as core values. In addition, Moodle aligns with regulatory frameworks such as the EU AI Act, the OECD AI Principles, and the NIST AI Risk Management Framework.
Moodles AI Principles: The Ethical Framework
The technical implementation follows clearly defined principles:
- Transparency: Users always know when they are interacting with AI. Content is labeled as AI-generated.
- Configurability: Institutions decide for themselves whether, where, and how to use AI. Moodle’s modular architecture makes this possible.
- Privacy and Security: Learner data is protected. Because open-source models can be run locally, sensitive data can remain entirely within your own data center.
- Human-centered approach: AI supports the learning process but does not replace pedagogical decisions. Humans remain in control.
Conclusion
Moodle demonstrates that integrating AI into e-learning doesn’t have to be an either/or choice between innovation and ethics. The AI subsystem in the Core is well-designed, modular, and gives institutions full control. It doesn’t lock any organization into a single provider, respects data protection requirements, and prioritizes pedagogical decisions over technical capabilities. For anyone using Moodle, it’s worth trying out the AI features—with the assurance that control always remains with the institution and the instructors.
