Editorial planning
SEO editorial plan: approval and AI observations
A draft is ready and the editorialPlan page shows observations from Google and AI engines. Here is how human approval separates what was observed from what still needs checking before publishing.
You have a draft ready, and the editorialPlan page shows observations from Google and from AI engines. The question is a practical one: how does approving an SEO editorial plan help you separate AI engine observations from what still needs checking before you publish? By default, we require human approval of a version. No draft goes live just because an observation looks favorable. That is exactly what approval is for: separating what has been observed from what has been checked. Automatic publishing is turned on per site and still goes through the checks, so it does not replace the verification step. On the editorialPlan page, a draft is approved only after you have confirmed which observations concern Google and which concern AI engines, keeping in mind that Gemini and Google AI Overviews are not the same observation.
Set up the editorialPlan page as the approval point
RankAI's working proposal is to treat the editorialPlan page not as a store of notes but as the place where a draft moves from "observed" to "approved." Observations collected from Google and AI engines remain reading material: on their own they are not a green light. Human approval of a version is the step that closes the loop, and it should happen only once every observation has been classified.
The observations we can name
The public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI. Claude and Grok are measured only after you connect them. Gemini and Google AI Overviews are not the same observation: keep them separate even when they address the same topic. That is the scope within which you should read what appears on the editorialPlan page.
Three steps to approve a draft without confusing observation and verification
- Classify each observation by engine. Verifiable criterion: every row on the editorialPlan page states the engine it refers to, and Gemini and Google AI Overviews appear as separate entries. If an observation cannot be attributed to an engine, it is not yet a usable observation.
- Mark the measurement status. Verifiable criterion: for Claude and Grok, the draft states whether the connection has been made; until it has, the observation does not exist and cannot support approval. For the other named engines, the observation is the one available on the page.
- Separate what still needs checking. Verifiable criterion: every point that does not come from a recorded observation is written as an open check, with an owner and a closing condition. The draft is approved only when the list of open checks is empty or explicitly accepted.
- Approve the version, not the idea. Verifiable criterion: human approval is recorded against the specific version of the draft. If the text changes after approval, the version changes and the approval does not carry over.
Why automatic publishing does not settle the question
Automatic publishing is turned on per site and still goes through the checks. That means human approval of a version remains the default behavior, and automation does not turn an observation into a verification. Even with automatic publishing on, the editorialPlan page still has to keep the distinction between what has been observed and what still needs checking.
The point that often gets missed is that an observation from an AI engine describes what that engine returned at a given moment, not what the page must contain. In our working proposal, the editorialPlan page should have two conceptual columns: observations on one side, with engine and date; open checks on the other, with owner and closing condition. Approval is given only when the second column is empty or explicitly accepted. A second precaution concerns cross-reading: an observation from Gemini should not be used to justify a choice about Google AI Overviews, because they are different observations. A third precaution is version traceability: if the text changes after approval, the draft goes back into verification. These precautions are not a universal standard but a method that makes the difference between what has been seen and what has been checked visible on the editorialPlan page.
A limit to keep in mind
The method described here does not guarantee that AI engine observations are complete or representative: it only describes how to organize them on the editorialPlan page before approval. An API response is not a person's screen, and that is worth remembering every time you read an observation.
For the full picture on how to set up the plan and approval, see the in-depth page.