AI quality check¶
The AI quality check is an automatic diagnosis of your course based on five didactic dimensions. In 15-30 seconds you get an overall score plus concrete improvement recommendations — ideal as a sanity check before publishing or when completion rates are low.
What the check evaluates¶
Five dimensions, each with its own score from 0-100:
| Dimension | What is checked |
|---|---|
| Completeness | Is all essential content there? Are there obvious gaps? |
| Structure | Is the setup logical? Is there a sensible common thread? |
| Content | Is the content professionally relevant, up to date and suited to the target group? |
| Learning objectives | Are the learning objectives clear and verifiable? |
| Difficulty level | Does the level match the declared target group? |
The five dimensions produce an overall score of 0-100 with color coding:
- 75-100 — good quality
- 50-74 — room for improvement
- 0-49 — considerable room for improvement
Starting the check¶
You can trigger the check in two places:
From the course catalog¶
- Find your course in the course catalog.
- Click the quality icon on the card.
- In the dialog, click Check now — or Check again if results already exist.
From trainer analytics¶
- Analytics → Courses.
- In the course table, the AI quality column shows the current score.
- Clicking the score opens the same dialog → Check again.
Screenshot to follow
Quality dialog with a doughnut chart for the overall score and dimension bars
While the analysis runs, a loading indicator appears ("AI is analyzing the course…"). If you close the dialog, the analysis is not aborted server-side — but the value only updates the next time you open it.
Understanding the results¶
In the dialog you see:
- Overall score as a doughnut chart + date of the last check
- Dimension bars with a score and individual feedback per dimension
- Strengths — a list of what works particularly well (a plus for your course!)
- Recommendations — concrete suggested actions, e.g. "add quizzes to check knowledge" or "formulate the learning objectives in lesson 3 more precisely"
What the check does NOT evaluate¶
The check analyzes the metadata and structure of the course — course title, description, difficulty level, lesson titles and lesson types. It does not look at the media quality of your content: videos, embedded H5P content, PDF attachments are not evaluated for their content. Judging media quality remains editorial — the AI is a tool here, not a replacement.
When a check is worthwhile¶
- Before publishing — a final sanity check that the structure and coverage are right
- After major content changes — a new block of lessons, adjusted objectives, added quizzes
- At low completion rates or poor learner ratings — as a diagnostic tool for what might be causing it
- Before an annual refresh — does the difficulty still match the target group?
There's no cooldown — you can start the check as often as you like.
Interpreting concrete recommendations¶
Examples and how to deal with them:
| Recommendation | What you can do |
|---|---|
| "Formulate learning objectives more precisely" | Reconsider: what should learners concretely be able to do after the course? At most 3-5 clear sentences in the description. |
| "Add more knowledge checks" | Insert quiz lesson(s) between longer stretches of explanation. See Quizzes and grading. |
| "Review the difficulty progression" | Are the first lessons really for beginners? Or do you jump into detail too quickly? |
| "Common thread unclear" | Check the lesson order, perhaps add a transitional text between topics. |
| "Estimated duration seems too short/too long" | Play through it yourself or ask a test learning group how long they actually need. |
Frequently asked questions¶
The check gives me a low score, but my learners are satisfied. Both views are valid — the AI evaluates didactic structure, your learners evaluate practical usefulness. If the course works in practice, the AI recommendations are an optional improvement, not a must.
How often should I check? During the creation phase, a few times, with a new check after each adjustment. In live operation, at most every few months. Frequent re-checks without changes deliver almost identical results.
My course has very different scores across several dimensions — what should I do? Focus on the lowest scores first — that's where you get the biggest gain per effort. Don't ignore the high scores, but setting priorities is the fastest route to a noticeable quality improvement.
The AI gives me recommendations I think are wrong. You know your course and your learners better than the AI. See recommendations as suggestions, not instructions. If a recommendation seems wrong to you, skip it.
Do learners see the score? Currently yes — if the catalog is configured to display quality scores. Low scores can be off-putting. If you want to improve first, check the configuration with your platform manager.