Validation and playtesting
Validation and playtesting reveal different kinds of problems. The product page's current requirements determine which action is available next.
Automated validation
Validation checks project structure, contracts and rules that the platform can verify deterministically. On failure, have the assistant inspect diagnostics, repair the current working content and validate again.
Passing validation does not prove that a product is enjoyable or understandable.
Preview and live playtesting
Preview checks a limited execution context. Live playtesting runs the actual flow against specific running content and records evidence for it.
Use the playtest entry when available, choose the mode to inspect and complete the main path. After changing the content, earlier tests do not prove that the new content works. Validate and test the affected paths again.
AI and human evidence
| Method | Useful observations |
|---|---|
| AI playtest | Legal actions, reachable paths, completion and machine-observable errors |
| Human playtest | Clear instructions, reasonable waits, trust and enjoyable interaction |
The platform distinguishes tester types. AI completion is not a human experience review. Do not turn suggested test coverage into requirements that the publication page does not impose.
Useful feedback
Record the mode, steps, expected result and actual result. For example: “After the second player submits an answer, neither player has a next action.”
Ask the assistant to repair the original product, then recheck the affected paths. Review current service costs before testing; private playtesting does not imply every external service is free.