Editorial planning
SEO editorial plan: approving with AI observations
How to use RankAI observations from Google, ChatGPT, Perplexity, Gemini, Grok and Google AI to approve an SEO editorial plan without treating them as a guarantee of ranking.
If you are building an SEO editorial plan with approval, the practical question is this: how do I use RankAI observations on Google, ChatGPT, Perplexity, Gemini, Grok and Google AI to decide which content to approve, without treating those observations as a promise of ranking? Our starting point is a workflow in which a person approves a version before it is published. The observations inform that decision; they do not replace it.
What we observe and what stays out
RankAI's public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI. Claude and Grok are measured only after you connect them. We keep two things separate that are often confused: Gemini and Google AI Overviews are not the same observation. An observation collected through an API is also not the screen a person sees: it is a product data point, not a snapshot of the user experience.
These are limits of our tool, not guarantees. An observation describes how an engine answered in a given context; it does not say that content will earn traffic or positions.
A three-step approval method
We offer this sequence as a RankAI operating method, not as a universal standard. Each step has a verifiable criterion you can check before you approve.
- Collect observations per engine, separately. Verifiable criterion: every row in the plan states which engine was observed (ChatGPT, Perplexity, Gemini, Grok, Google AI) and whether the observation is Gemini or Google AI Overviews. If two observations are labeled as the same engine but come from different sources, the row is not approvable.
- Connect the engines that require a connection before counting them. Verifiable criterion: rows for Claude and Grok appear in the plan only after you connect them; until they are connected, they stay out of the count of usable observations.
- Approve a version, not an idea. Verifiable criterion: approval applies to a specific version of the content, with a date and a reference to that version. If the version changes, the previous approval no longer holds.
- Treat automatic publishing as a controlled exception. Verifiable criterion: automatic publishing is turned on per site and still goes through the checks; the default remains human approval of a version.
Where observation ends and the decision begins
An observation can guide the choice of a topic, an angle or a wording. It cannot say whether that content will work. The decision to approve stays with a person, and the editorial plan should make visible who approved what and on which version. That way, if content does not produce the expected results, you can reconstruct the path: which observations were available, which engines were connected, which version was approved.
The contribution this piece adds is a criterion for separating observation from approval, written for people drafting an editorial plan. The proposal is to treat every row in the plan as a card with four required fields: engine observed, type of observation (distinguishing Gemini from Google AI Overviews), connection status for engines that require it, and approved version with date. A row that does not fill all four fields does not enter the approved plan. This does not improve SEO results by construction: it only makes traceable why a piece of content was approved. The second part of the contribution concerns automatic publishing: because it is turned on per site and still goes through the checks, it is worth recording in the plan which sites have automation active, so that human approval remains the default behavior and the exception is explicit. Finally, the plan should state that an observation via API is not the same as a person's screen: whoever approves must know they are reading a product data point, not a real user's experience.
A verifiable limit
Before you consider the plan closed, check one thing only: for every approved piece of content, there is an identified version and an observation labeled by engine, with Gemini and Google AI Overviews kept separate. If either element is missing, the plan is not yet approvable under this method. Observations remain an input, not a guarantee of results.
For the full picture on the editorial plan, see the in-depth page.