Editorial planning

What to Check Before You Approve a Version

Five checks decide whether a version is approved: brief, facts, editorial plan, tone and structure, and no unverifiable promises. A version that misses one goes back for revision.

Pico NeuralOctober 22, 2026

Before approving a version, we run five checks: it answers the brief, its facts can be verified, it fits the editorial plan, tone and structure are consistent, and it makes no unverifiable promises. If any check fails, the version goes back for revision. Human approval is the default in RankAI. Automatic publishing is turned on per site and still goes through the checks. The checklist below is our working proposal, not a universal standard.

The approval checklist, item by item

Take an article that is ready for an SEO editorial plan with approval. Before signing off, the reviewer works through the five items and ticks them one by one.

1. Does it match the brief?

Does the version answer the reader question set out in the brief? If the brief asked for a procedure and the version only gives background, the item stays unticked.

2. Are the facts verified?

Every claim about the product is one we can support. No numbers, prices or promises that are not in the facts register.

3. Does it fit the editorial plan?

Does the version fill the slot the plan gives it, and does it avoid duplicating another piece? The test is checkable: if two articles answer the same question in the same way, one of them needs rethinking.

4. Tone and structure

The answer comes in the first lines, headings are H2 and H3, and the text speaks as “we” or as RankAI.

5. No unverifiable promises

There are no guarantees about rankings, time savings or results that are not documented.

A version that passes

An article for the editorial plan comes up for review. It opens with the direct answer, cites only facts from the register, does not duplicate a published piece, uses the right headings and promises no results. The reviewer ticks all five items and approves it. If item 2 stays unticked because a figure is not in the register, the version goes back for revision with the exact line to fix.

Our proposal is to treat the checklist as a gate, not a traffic light: each item gets a yes or no, and the version passes only with five yeses. That avoids two common mistakes. The first is approving “almost”: a version with one grey item is sent back, not approved with reservations. The second is confusing content review with publishing checks: the checklist is about the text, while any automatic publishing is a per-site setting that still goes through the checks and does not replace human review. In practice, the reviewer fills in the checklist before looking at the preview, so the judgement is about the text, not the rendering. Keep the checklist next to the approved version: if a fact changes later, we know which item to reopen.

One limit we can verify: on the technical trial the limits are 3 articles per month and 3 in total, so the checklist applies to a small number of versions and we cannot draw conclusions about higher volumes. For the editorial plan context, see the in-depth page.

All articles

Query, language and market: three separate choices

Dates and priorities in an SEO editorial plan: keep them separate