# Structured AI Edits Need Platform-Native Validation ## Claim An AI editing tool should validate structured content through the platform's real render path before declaring the edit successful. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - This applies when content has platform-specific structure beyond syntactically valid HTML, such as WordPress block markup, an AST, or schema-bound documents. - The check confirms compatibility with the platform's interpreter; it doesn't prove the edit matches user intent or produces an accessible design. - It weakens when the render path itself is under modification or too slow to run at edit granularity. ## Evidence [[3. Resources/Readwise/Articles/More Targeted Content Editing With Block MCP]] describes Block MCP re-rendering every edit through WordPress to confirm that WordPress interprets the block structure as expected, paired with a revision that permits rollback. ## Links - [[Watching a System Run Beats Reading It]] — applies-in: rendering the changed blocks provides runtime evidence that inspecting plausible markup cannot. - [[Encoding What Good Looks Like Is What Makes Agent Verification Trustworthy]] — extends: the platform interpreter supplies one concrete, executable definition of structural validity. - [[A Revision Is Legitimate in Proportion to Its Auditability]] — because per-edit revisions preserve a recoverable record when validation misses an intent-level failure. - [[3. Resources/Readwise/Articles/Contributors Call for Testing Responsive Styling Ahead of WordPress 7.1 After Years of Community Demand]] — evidence that platform behavior needs contributor testing in the real editor and render path. - [[3. Resources/MOCs/AI Agent Safety MOC]] — applies-in: the MOC lists this note under Synthesis addition — 2026-08-11.