Beyond Authoring Tools: How We Built a SCORM-Compliant Quiz with AI

In the world of e-learning, creating custom, interactive, and trackable content has traditionally meant one thing: expensive authoring tools. Software like Articulate Storyline or Adobe Captivate are the industry titans, offering powerful features but often at the cost of steep learning curves, restrictive licenses, and "black box" outputs.

But what if you could build a fully custom, SCORM-compliant assessment without any of them? What if your only tools were a clear idea and a conversational AI partner?

That's exactly what this article is about. The 8th-grade science quiz we just built wasn't created with an authoring tool. It was built from scratch, line by line, through a collaborative process between a human and an AI.

Here’s how we did it, and why this approach represents a significant shift in content creation.

The Process: From an Idea to an Interactive Quiz

Our development process was rapid, iterative, and conversational. We didn't follow a traditional, linear development cycle; we built and refined in real-time.

Step 1: The Core Challenge (The "SCORM Problem")

The project started with a simple question: "Is it possible to build an HTML interactivity in Gemini that would make SCORM calls?"

This is the foundational technical hurdle. Instead of spending hours reading SCORM documentation, the AI was tasked with generating the necessary JavaScript "wrapper." This wrapper's entire job is to find and talk to the Learning Management System (LMS) API. After a few rounds of debugging (including the classic "API not found" and window.opener issues), we had a robust, working script that could Initialize, SetValue, Commit, and Terminate—the four pillars of SCORM communication.

Step 2: The Content (AI as Instructional Designer)

With the technical foundation in place, the request changed: "Let's build a really engaging assessment... 10 questions... 8th-grade science."

The AI immediately shifted roles from SCORM-expert to instructional designer. It generated:

• The Content: Ten 8th-grade-level science questions.
• The Interactivity: A wide variety of question types, including Multiple Choice, True/False, Fill-in-the-Blank, Select-All, Drag-and-Order, and Drag-and-Match.
• The Data Structure: A clean JavaScript questions array to hold all this data, separating the content from the presentation logic.

Step 3: The Interface (AI as Developer & Designer)

A list of questions isn't an assessment. The next request was to make it "colorful and fun." The AI acted as a front-end developer, writing all the HTML, CSS (using Tailwind), and JavaScript to:

• Create a "Start" screen, "Quiz" screen, and "Results" screen.
• Render one question at a time.
• Dynamically build the different interaction types based on the questions array.
• Add a progress bar, fun button styles, and a responsive, mobile-friendly layout.
• Add a final "Review Answers" screen, complete with logic to show the user's answer alongside the correct one.

Step 4: The Deployment (AI as Technical Writer)

An HTML file alone can't be uploaded to an LMS. When asked "can you generate the imsmanifest.xml file," the AI produced the complete, valid XML file. It explained why this file was necessary: to tell the LMS that this package was a SCORM 2004 course and that scorm_test.html was the file to launch.

Step 5: The Iteration (AI as Debugger)

This is perhaps the most powerful part of the process. When bugs were found, there was no need to dig through forums. The user simply described the problem:

• Bug: "Q4 is not tracking my selections."
  AI Fix: The AI identified a bug in the grading logic for the drag-order question and provided the exact code change.
• Bug: "Q7 marked me wrong, but my answer was right."
  AI Fix: The AI explained why (a "string-compare" error on objects) and wrote a new compareObjects function to fix it immediately.

The Benefits of Building with AI

This "conversational development" workflow has profound benefits over the traditional authoring tool model.

• Unprecedented Speed and Iteration: We went from a simple question to a fully functional, SCORM-compliant, 10-question interactive assessment in a single session. Adding a major feature like the "Review Answers" screen took minutes, not days.
• Total Customization (No "Black Box"): We are not limited by templates. Every line of code in the scorm_test.html file is open, commented, and editable. We built custom drag-and-drop logic from scratch. If you can imagine a new interaction type, the AI can help you build it. This is a level of freedom impossible to achieve in most authoring tools.
• Cost-Effectiveness and Accessibility: This entire process was free. There were no $1,400/year subscription licenses for software. This democratizes content creation, allowing anyone—teachers, trainers, or small businesses—to build professional, custom e-learning without a large budget.
• Integrated Expertise on Demand: To complete this project, we needed a SCORM expert, an instructional designer, a JavaScript developer, and a QA tester. The AI seamlessly transitioned between all these roles, acting as a collaborative partner that filled in the technical gaps, allowing the user to focus on the end-goal.

Conclusion

The assessment we built is more than just an HTML file. It's a proof of concept for a new way of working. This AI-driven approach frees creators from the high cost and creative constraints of traditional authoring tools, enabling a future where custom, engaging, and effective learning content can be built by anyone, anytime, at the speed of conversation.