TL;DR
Different evidence can support different decisions at the same meeting. Correct a verified public error, investigate an unresolved observation and hold a proposal that lacks support. Use a decision record so a changing AI mention count does not automatically become a request for more content.
Start with the review packet you actually have
Confirm the agreed service, location and observation method before discussing results. Open the dated answers, cited pages and operational fact record behind each proposed action. Use the AI visibility report specification to identify missing evidence.
Ask the owner to separate decisions that can proceed from conclusions that remain uncertain. An error in a controlled page and an unresolved answer elsewhere do not need to share the same status. Apply the decision worksheet to the actual evidence in your review.
Keep comparable and exploratory observations apart
Compare the unchanged core panel across review periods. Record new exploratory questions separately. Combining the original panel with new questions changes the group being measured, so the combined result cannot substitute for the original comparison.
Keep the two groups separate. A change within a small core sample does not establish a market-wide trend or its cause. The exploratory results may suggest useful questions, but they cannot retroactively replace unfavorable core observations.
Google Search Console’s Performance report defines search metrics such as clicks and impressions for its own reporting context. Google’s Performance documentation. Do not relabel a manually sampled mention as a Search Console impression or treat either as an attended appointment.
Choose the action the evidence supports
| Condition verified in the review | Owner decision | Who acts and what closes the item |
|---|---|---|
| A controlled page contradicts an operational fact the responsible person has confirmed | Correct the demonstrated error | Editor prepares accurate wording; implementer checks the rendered page and affected links |
| A current source is correct but a saved assistant answer changes its meaning | Keep the answer observation open | Measurement owner preserves the discrepancy, uses available feedback where appropriate and repeats the declared observation method |
| New questions have been added to the original comparison group | Pause the combined performance conclusion | Reporting owner separates stable and exploratory observations and identifies any remaining limits |
| A content proposal rests on changed mention counts without a demonstrated reader need | Return the proposal for a content-gap explanation | Editorial owner checks unanswered questions and their existing homes before commissioning work |
These conditions may occur together. Correcting a verified page error does not require accepting a broader growth claim. Use the content-map method when the unresolved decision is whether more articles are needed.
Complete a decision record for each item:
| Field | Entry from this review |
|---|---|
| Issue and supporting evidence | _____ |
| Decision and its limits | _____ |
| Responsible owner | _____ |
| Next action or dependency | _____ |
| Evidence needed to close or reconsider it | _____ |
Record what the decision does not establish
Verifying a repaired page does not prove every assistant refreshed. A later correct answer does not prove the correction caused it. Declining the expansion does not mean content can never help; it means this proposal has not shown why the proposed pieces are needed.
Ask for understandable evidence and decisions in the agreed reporting format. The issue, its owner and its closeout condition should remain clear whether delivered in a document or through software.
Carry unresolved items forward intact
Use the same issue ID at the next review. Show what changed, which dependency remains and whether the decision should be reopened. Do not rename an old unresolved answer as a new finding every month.
If the owner chooses to broaden the pilot to another service or location, record a deliberate scope change and establish the new comparison method. Expansion is a business decision, not a way to improve the appearance of the old series.
Frequently asked questions
Does uncertainty mean we should wait on every action?
No. A verified wrong office fact can justify correction while a broader performance conclusion remains uncertain.
What if the next observation improves?
Retain it alongside the earlier evidence. Review whether the method and context stayed comparable before claiming a trend or cause.