Use real screen recording for product operations and label AI examples according to their actual role. A previously generated artwork can illustrate a possible result, but it should not appear to be the output of a task you did not run in the tutorial.
Sort the material before editing
Create three groups: current interface recording, existing generated examples and conceptual workflow illustrations. Record the source and permissions for each. Hide private account details, balances, email addresses and unrelated user assets before including a screen recording.
In Imagild, capture the actual entry point and controls relevant to the tutorial. If Film Studio is used to plan supporting scenes, keep those scenes separate from interface evidence. A generated button animation does not prove that a product function exists.
Explain the example honestly
A fictional tutorial demonstrates choosing a Skill and adding an instruction. It then shows an existing authorized output. Label that output as a previous example instead of implying that the recorded selection produced it immediately.
If recording a real generation, show the verified result and disclose any shortened waiting period. Do not convert edited runtime into a measured speed claim. If the task fails, either explain the failure or make a different truthful demonstration; do not splice in an unrelated success.
Check narration and visual evidence together
For every claim, identify the shot that supports it. Ensure the narration does not say “your image is ready” over an unverified example. Keep limitations visible or spoken when they affect the viewer's decision.
Review the complete export for private information, readable subtitles and accurate labels. Save a source log and the final script. This makes later updates easier when the interface changes, and prevents a conceptual illustration from being reused as factual product evidence.