CORE GUIDE
Media quality review
Media quality review evaluates generated images, audio, video and presentation assets against instructional purpose, factual accuracy, accessibility, brand constraints and technical delivery rather than accepting output because it looks polished.
Mental model
Generated media is a product artifact with a job to do. Review should ask whether the asset communicates the intended concept correctly and survives the real viewing or listening environment, not only whether it is aesthetically pleasing.
Why it matters
AI can create visually convincing media that contains wrong labels, unreadable text, inconsistent diagrams or accessibility failures. In educational products those errors can teach the wrong mental model. A review contract turns taste into inspectable acceptance criteria.
01
Review meaning, usability and delivery separately
Start from the learning objective and list the information the asset must convey. Check semantic correctness, hierarchy, legibility, captions or transcripts, color/contrast and device constraints. Then validate technical properties such as dimensions, duration, compression and file size. Keep a reject/revise reason so repeated generation failures become data for improving the production workflow.
02
Example: beautiful diagram teaches the wrong data flow
A generated architecture illustration is visually strong but places the verifier before the tool result, implying the system verifies intent rather than outcome. A semantic review catches the causal error, while a purely aesthetic review would approve it. The asset is regenerated with the correct flow and then tested at mobile size for label legibility.
Common failure modes
- Approving generated media primarily on visual appeal.
- Checking technical file properties while missing a wrong concept or label.
- Treating accessibility fixes as optional polish after publication.
Engineering heuristics
- Tie each asset to an explicit communication or learning objective.
- Review semantic correctness before aesthetics and compression.
- Test the final media in the actual device, caption and accessibility context.
Takeaways
- 01Polished media can still teach the wrong thing.
- 02Quality review needs semantic, accessibility and technical gates.
- 03Reject reasons should feed the next generation or editing step.
Reading evidence
This records actions you actually took; it does not claim mastery, proficiency, or certification.
Used in
This Concept is reused across these canonical learning paths.