How AI is used
Which parts are written by a model, which are not, and what that means for you.
Every factual statement here is drawn from what the software actually does, and the text has been reviewed. It is still a description of a service rather than legal advice.
What the model writes
The course itself. A model reads your brief and settles what the course is about; a second designs the curriculum; a third plans the slides and chooses the course’s colours, fonts and visual style; and one model per slide designs and writes that slide — its layout, its wording, its activity and the code behind it. Any photographs are generated too.
The voice, when a course has one
A course built with voice-over is narrated by a synthetic voice, not a person. A model writes what is said for each slide, and a speech provider (ElevenLabs) reads it aloud in the voice the course’s creator chose. The same caution applies as to the rest of the course: the words are a draft to be read before learners hear them, and the subtitles show exactly what is being said.
What the model does not write
Anything your LMS depends on. The SCORM manifest, the run-time wrapper, and the course frame that carries navigation, progress, resume and every value reported to your LMS are written by code and are identical in every course. A slide reports through that frame; it never talks to the LMS itself.
This is the reason a generated course cannot forget to report its score. It is not a matter of prompting the model carefully; that part is simply not the model’s to get wrong.
What is checked automatically
Before you can download a package, the manifest is cross-checked against the files actually present, and every slide is opened in a real browser against a simulated LMS: that it loads without errors, that its text fits, and — where a slide is gated or scored — that it can be completed and that the score and each answer reach the LMS. Screenshots of every slide are also reviewed for visible defects.
These checks catch structural and visual failures. They do not, and cannot, judge whether the teaching is any good — that judgement is yours.
Review before use
A generated course should be read before it reaches learners. Models produce confident, fluent text that can still be factually wrong, and a course carries more authority with a learner than a chat reply does.