Review the first month of an AI visibility engagement against the agreed work and the evidence for each item. Open the baseline, check what shipped, and ask the receiving team whether it can use what was handed over. Then record what you accept, what needs correction and what will carry into the next month with a named owner.
Keep discovery results in their own part of the review. An answer may have changed before a page went live. A change that went live yesterday may have no meaningful follow-up yet. The review should show which work was delivered and which observations the evidence can support at that point.
AI Visibility is published by Lunasol, which offers GEO and AEO services. The month-one example below is fictional. Public sources and tool documentation were checked on 5 November 2026. AI Visibility publication disclosure
Reconcile the report with the scope
Bring the latest agreed scope, any approved changes and the agency’s delivery record into the same review. Start with the items due this month. For each, note the promised output, the evidence supplied and the person who can confirm it. If the scope was revised, keep the approval attached so the team can distinguish a deferred item from work the agency still owes this month.
The existing 90-day plan describes how a business might sequence AI visibility work. The monthly reporting guide covers what to include in the report. Here the report is one input to a buyer decision. A useful review can accept one item, return another and keep a third open for an observation date. AI Visibility guides and monthly reporting guide
Ask the agency to indicate which work was completed by its team and which depended on the client. If the writer delivered a draft and the client still needs to approve a service fact, record the draft and the open approval separately. If the agency promised publication but the release slot moved, identify when the change was agreed and what work can continue while it waits.
Verify a sample of the actual work
Select one item that was meant to be released, one piece of research evidence and one open dependency. Ask the person who owns each item to show the source record. A release can have a dated page address, before and after evidence and a check of the relevant customer action. A research finding can link to an observation and the source behind it. A dependency can have an owner, a request date and a due date.
For a live page change, open the affected page and check the approved fact or destination. A draft, a staging view and a public page support different conclusions. If the agency planned to revise four pages, check which addresses actually carry the change. Look for a record of the remaining pages rather than assuming all four inherited it.
009 Agency’s public INC4 case separates dated observations from work it reports shipping. It also says the client’s developers deployed the site changes and describes gaps found when reported work was compared with production. The case illustrates why a reviewer should ask what was checked live and who did the check. The public article gives the agency’s account of the work. 009 Agency’s public INC4 case
If one sample fails, decide whether it indicates an isolated exception or a wider issue. Ask for evidence from another affected page before expanding the review. Keep the additional verification targeted so a narrow discrepancy does not turn the meeting into a fresh audit.
Check whether the evidence can support a comparison
Open the agreed baseline and compare its question set and conditions with the latest observations. Identify any questions that changed wording, any new markets and any surfaces that were unavailable. The agency can include a new observation alongside the baseline if it marks the change clearly. A comparison should preserve which entries were measured under comparable conditions.
Google’s official 13 October 2026 model announcement describes different product and subscription contexts in the same caption. In a first-month review, a new model option can prompt a useful discussion about what was measured. Ask whether the agency used the same experience both times and record any difference before using the numbers to make a decision. Google’s official model announcement
Review first-party numbers with their report settings. Google’s Search Console generative AI performance report for Search includes impressions for AI Overviews and AI Mode. Its help page says recent figures can be preliminary and that the exported file can turn unavailable values shown as symbols into zeros. Ask the agency to preserve the report view or note which zeros represent unavailable values. A spreadsheet total alone can hide that distinction. Search Console generative AI performance report
Record the report period, property and relevant filters with the file. If the business compares platform reporting with a Dubai release date, note that Google documents the report dates in Pacific Time. Agree a period suitable for the question before interpreting short windows around a release. Search Console report dimensions
Make a decision for every item due
Imagine a fictional first month with four agreed outputs. The buyer expected a baseline, two page updates and a follow-up check after release. One update went live. The other is awaiting the client’s approval of an operating condition. The analyst attempted a follow-up on a surface whose access conditions differed from the original run.
| Agreed output | Evidence at the review | Buyer decision |
|---|---|---|
| Baseline for the approved questions | Dated record and usable export with run conditions | Accept the baseline |
| First page update | Public page shows the approved change and the intended contact route opens | Accept the release |
| Second page update | Draft exists and a service fact awaits client approval | Keep the draft recorded and assign the approval and release dates |
| Follow-up observation | Recorded result under changed conditions | Retain the observation and schedule a comparable check where possible |
In this example, the follow-up record can be complete as a record of what happened while the intended comparison remains open. The buyer can ask the analyst to retry under comparable conditions or agree on a new baseline if the original experience is no longer available. The decision depends on the agreed scope and what the team can actually verify.
The second page needs its own decision. If the agency’s responsibility ends at delivering a draft, the buyer can accept that stage when it meets the brief. If the agreed output was a live page, leave the release open and record the reason. The agency and client should be able to explain the same status from the same document.
Turn a discrepancy into a next action
If the report says a change was published and the reviewer still sees the previous text, capture both records with their dates. Ask the release owner to check the relevant address and environment. A staging link, a different language version or a cached view can explain why two reviewers saw different things. Confirm which public version is in scope before closing the issue.
Agree on a specific response. The agency might provide a missing reference, correct the deliverable or explain why a dependency prevented completion. The buyer might approve a revised date or supply an outstanding fact. Set a next check date and use the same evidence standard when the item returns.
Keep an unresolved commercial or operating decision with the person who can make it. Suppose a service area changed after kickoff and the website team has two competing lists. The agency can show the conflict, but the business must settle the current coverage before it can accept wording based on either list. Record the decision and make sure the next month’s tasks use the approved version.
Check whether the first month leaves a usable handover
Ask the internal team to repeat one step without the agency guiding the screen. The website owner can locate the published change, the analyst can open a saved observation, and the service owner can find the facts still waiting for approval. This is a practical check on whether the team can use the material it received.
For example, imagine the agency shows a baseline dashboard and agrees that the buyer should receive an export. The reviewer downloads the sample and notices that it contains counts but omits the exact questions and dated conditions. The presentation may still be informative. The agreed evidence handover needs a correction if the missing fields were part of the scope. Give the agency the specific fields and a delivery date, then open the replacement file before closing the item.
A similar distinction applies to recommendations. If the agreed handover is a developer task, ask the person who will do the work whether the record identifies a page, a reproducible issue and an expected result. An analyst’s note can correctly identify a problem while leaving implementation details for a later task. Decide whether that division matches what was commissioned.
Treat changes in scope as decisions for the next plan. If the agency proposes a new market after the first report, ask what existing work would move and what evidence supports the expansion. Record the new option separately until it is approved. A month-one review can surface worthwhile additional work while still closing the original commitments on their own terms.
One helpful final question is what the agency would need from the buyer in the next seven days. Ask for one named person and a due date for each input. The buyer can then accept the plan based on resources it can actually provide.
Decide what to continue in month two
Use the review to choose the next work. If the baseline is usable and the main release has happened, continue with the agreed verification period. If evidence is missing, reserve time to repair it before extending the test set. If several tasks depend on the same overdue approval, give the decision owner a deadline and move work that can proceed independently.
Check the intended measurement window against what actually shipped. A page updated near the end of the month may need a later observation. Record the release date and the earliest review date agreed for that change. If a recent platform report is preliminary or unavailable, show that status and name the person who will verify it.
Write the result of the review so the team can act without replaying the meeting. Include the accepted items, the items returned for correction, dependencies the client must resolve and observations that need another check. Identify any decision about scope and who approved it. A short record of decisions is more useful than a second slide deck repeating the original report.
Before the next review, confirm that the buyer can reopen the underlying files and find the evidence for any closed item. The next month can then start from a clear delivery record, with open tasks and follow-up observations assigned to the people who can complete them.



