Answers
When Your CMS Won't Publish: What to Do in RankAI
If your CMS won't publish directly, don't guess why. Check what RankAI actually shows, read N/A correctly, and log what you saw.
If your CMS won't publish directly, there's no need to guess why. Work with the connector limit using what RankAI can actually show you. Open the page in question and check whether the visibility figure is there. If it's missing, RankAI displays it as N/A, never as zero. That's the first fixed point: a missing figure is not treated as a negative result. From there, separate the single case from the recurring one, and don't attribute behavior to the connector that you haven't observed. If checkout is off during pre-launch, the invitation is to create an account or open the relevant page, not to start an immediate trial.
A decision tree for the connector
Our approach is a decision tree, not a threshold. The question isn't "how many times did this happen" but "what can I verify right now." That keeps a connector limit from being confused with a missing measurement.
First node: is the figure there?
If the visibility figure is missing, the page shows N/A. Don't read that N/A as proof the CMS blocked publishing. It's our missing-data rule being applied, not a recorded fact about how the connector behaved. N/A says one thing only: that figure isn't available.
Second node: single case or recurring?
If the problem involves one page, treat it as a single case and note what you saw. If the same behavior shows up across several pages, treat it as recurring and note the sequence. Either way, you record the same three things: the page, the figure shown (N/A included), and when you observed it.
Third node: what you can't conclude
A page in simplified Chinese does not mean RankAI measures 豆包, 元宝, Kimi or 千问. Likewise, an N/A doesn't let you conclude the connector is blocked. These are limits of the data, not diagnoses.
The contribution here is a two-branch decision tree with one tracking rule. Branch A, single case: the page shows N/A, you log it as a single case and don't generalize. Branch B, recurring case: several pages show N/A or the same behavior, you log them in sequence and treat them as recurring. Tracking is identical in both branches: page, figure shown, time. What separates them isn't a threshold but the fact that Branch B needs a sequence, not a count. This format is a RankAI working proposal. It doesn't replace a check on the connector itself. Its job is to head off two opposite mistakes: treating an N/A as zero, and treating a single case as recurring.
Two filled-in mini-cases
Single case. Page: product page. Figure shown: N/A. Time: when the page was opened. Reading: figure unavailable, no conclusion about the connector.
Recurring case. Pages: product page, category page, article page. Figure shown: N/A on all three. Time: the same review session. Reading: sequence logged, to be treated as recurring, without automatically pinning it on a CMS block.
The limit you can verify
Here's the limit: we can state that a missing figure is shown as N/A, and that the tree separates a single case from a recurring one. We can't state, from an N/A, that the CMS connector blocked publishing. That's why the check stays on the figure shown, not on the hypothesis.
For cases that don't fit this tree, see the FAQ page.