Content QA: One Pass for JSON-LD, FAQ and raw-md

What this helps you with

Before publishing you cannot tell whether the content is right: is JSON-LD complete, does FAQ match the page, can AI read the raw-md? Content QA is a pre-publish check: one piece of content, three views compared.

Run it in three steps

1. Open Content QA and choose a type (product / page / collection / article). 2. Enter the page path (e.g. /products/xxx) and click Run. 3. Read the result: PASS / WARN / FAIL, each with a reason and evidence location.

How to read the result

• PASS: the three views agree; safe to publish;
• WARN: risk exists (e.g. raw-md missing), not blocking;
• FAIL: hard error (e.g. missing Product schema); fix before publishing.

We never turn "unmeasured" into "passed": it is explicitly reported.

Want the rules? Read the expert version: content-qa.expert

Content QA (Expert): Rules, Template Differences and Automation

Chapter 1: Rules and template differences

1.1 Three views

1.2 Template awareness

• product: requires Product + Offer (price/currency/availability);
• page/collection/article: page-type Schema (FAQPage/CollectionPage/Article/Organization), not product rules;
• missing raw-md → RAWMD_FETCH_UNAVAILABLE (unmeasured, never fake PASS).

1.3 Result model

PASS / WARN / FAIL + reasons[] + evidence (field location / snippet / hash). Repeated runs on the same input should be reproducible (except timestamps/versions).

Chapter 2: Automation and troubleshooting

2.1 API

• Single run: POST /api/v1/content-qa/run (type + path)
• Batch: content-qa/batches (enumerated by type, queued)
• Results stored in content_qa_runs (mode / target_type)

2.2 Common FAILs

2.3 Relation to delivery

Content QA is a pre-gate in the delivery DAG: FAIL blocks publish; WARN needs human confirmation; only PASS goes to approval/publish/readback/rollback.

Back to beginner version:content-qa.novice