How do I read the monthly report and the workspace, and which numbers come first?

Start with the denominator and evidence completeness, then the cycle-by-cycle trend of mention rate and citation rate, and only then drill into single prompts to read the answer text, screenshot and sources. The order matters: without knowing how large the denominator is and whether the evidence is complete, a rise or fall in mention rate has no basis for interpretation. The workspace provides these figures per cycle; the formal monthly report is delivered as a PDF that can be circulated within your team.

Step one: denominator and evidence completeness

Each cycle's denominator is the number of unbranded prompt × platform observations that produced a normal answer, and it moves with the number of failed, blocked, refused and not-triggered tasks. Evidence completeness is the share of tasks with complete evidence; the runtime-exception rate is failed plus blocked divided by planned tasks. When one cycle's exception rate is clearly elevated, discount that cycle's metrics; when completeness is low across several cycles, contact us to investigate before discussing trends at all. All of these figures are listed per cycle in the workspace monitoring log:

Step two: the trend

Mention and citation rates are laid out by cycle, roughly 10–11 per month at 72-hour intervals against a frozen prompt set, so cycles compare directly. Treat the visibility score (mention rate × 0.7 plus ranking quality × 0.3) as a composite reference within one project, not as a cross-brand benchmark. Single-cycle swings are common; several consecutive cycles moving the same way are what deserve a root cause, such as a cited page being rewritten or a platform update.

Step three: drill into a prompt

Every movement in the trend must ultimately be traced to specific prompts. From a cycle, open a platform, then a single prompt, and you see the answer text, page screenshot, cited sources, status and stable task ID. When the brand is mentioned, check its position and the sentiment label; when it is not, see who the AI recommended instead and which pages it cited. Read the 6 branded fact checks separately, looking at whether the AI describes the brand's facts accurately; anything that cannot be verified is flagged as pending.

Handing source URLs to your content team

Cited sources are stored one by one with URL and title and are not deduplicated across prompts, so a page that appears under several prompts is recorded each time, which is precisely what shows its weight. Source URLs together with the prompt, platform and cycle they appeared in are structured output and can be supplied as CSV within the delivery scope; project and managed engagements also deliver a human-curated action list. The content team's job is then to judge whether those pages describe your brand accurately and to pursue corrections along the lines of /corrections-policy.

When is the monthly report issued?

The formal monthly report is a PDF covering every cycle in that subscription month. Workspace data becomes visible as tasks complete, so there is no need to wait for the PDF.

Mention rate went up. Can I put that in my management update?

Confirm first that the cycles involved have comparable denominators and evidence completeness. Miaowa GEO does not promise any uplift, so report the denominator, cycle count and exception rate alongside the rate itself.

Can colleagues without an account see the workspace?

Workspace access is granted per account, project and role. The PDF monthly report can be circulated within your team.

Last updated 2026-09-12. The capabilities page and the pricing page are authoritative where they differ.