How to Run a Pre-Publication Review (6-Dimension Rubric)

Why it matters

Most quality problems are cheapest to fix before publish, and a consistent rubric is what makes "good enough" objective rather than a feeling. The framework's pre-publication review scores every page 1–5 on six dimensions with hard thresholds, so a page with weak sourcing or high duplication risk is caught and blocked before it ships.

The rubric also produces a short review note — strongest feature, weakest feature, biggest unsupported claim, and the one change that would most improve citation-worthiness — which doubles as a record and a to-do list.

What good looks like

Every page passes a scored, recorded gate before it ships:

• The page is scored 1–5 on all six dimensions.
• It meets every minimum: Originality ≥ 3, Source quality ≥ 4, Provenance transparency ≥ 4, Structural coherence ≥ 4, Duplication risk ≤ 2, Speculative overreach ≤ 2.
• A review note is written (strongest/weakest feature, any unsupported claim, top improvement).
• Pages that fail any minimum are blocked until fixed.
• The gate is applied to every page, including the How-To Guides themselves.

Step by step

1. Score the six dimensions 1–5 Rate the page on Originality, Source quality, Provenance transparency, Structural coherence, Duplication risk, and Speculative overreach. Be honest — the rubric only works if the scores are real.
2. Check each score against its minimum Compare each score to its threshold (table below). A single failed minimum fails the page — these are gates, not averages.
3. Write the review note Capture four things: the strongest feature, the weakest feature, any major unsupported claim, and the one change that would most improve citation-worthiness.
4. Block or fix anything that fails If any minimum is missed, the page does not publish until the issue is fixed and re-scored. This is the non-negotiable part of the gate.
5. Record the gate before publishing Keep the scores and note with the page (or in your editorial log). For the newsletter, this is the same self-audit the framework-qa-auditor runs each issue; surfacing it builds a citable accountability trail.

An AI-assisted pre-publication review is a first pass, not a sign-off. Language models can approve confidently and miss factual errors, so keep a named human reviewer responsible for the final scores — the same principle as Phase 6.3’s citation monitoring.

Worked examples

Common mistakes