Press New course on the dashboard to open the wizard. It takes a few minutes of your input, then builds the course in the background while you carry on working - a dock at the bottom of the page shows progress.
Step by step#
- Name it and set the level. Subject, exam board and year group help the AI pitch the language. Tick the purpose (exam prep, catch-up, enrichment) so lessons are shaped to it.
- Give it material. Upload a PDF or document, paste notes, pick something from your library of earlier uploads, or use Find with AI to search the web for specification documents and past papers and import the useful ones.
- Review the outline. The wizard extracts topics and subtopics (with a checklist of what the syllabus says must be covered) and lets you rename, reorder or remove before anything is generated.
- Appearance. Choose a cover image from the gallery or let the wizard pick one, and a colour.
- Build. The course is queued. Each lesson is generated, audited for accuracy and ordered so teaching comes before questions. You get a notification when it's ready.
Sizes and limits#
Courses are capped at a sensible size so they stay teachable: up to 60 subtopics, a few lessons per subtopic and roughly eight to twelve cards per lesson. Bigger syllabuses are better split into two courses (for example, Paper 1 and Paper 2).
Each lesson generated costs credits from your monthly allowance; the wizard shows the estimate before you press Build. If you run out part-way, the build pauses and resumes when credits are topped up - nothing is lost.
Uploads never create a course by themselves. They land in your library, so you can build several courses from one specification or reuse a past paper next year.
Common questions#
Can I edit what the wizard made?
Everything. Open the course, click any card to edit text, answers, images or hints; add or delete cards; reorder lessons.
Can I build from scratch without AI?
Yes. Create an empty course and add lessons and cards by hand from the course page.
My PDF is a scan.
Scanned PDFs are OCR'd during import. Very faint scans may need re-exporting at a higher resolution.
