Is there a public sample report, and what real data does it show?

Yes. /report is a real monitoring project Miaowa GEO runs on its own brand, published read-only with the same dashboard components a client workspace uses. The interface is in Chinese. The project runs the managed-service configuration: ChatGPT, Gemini, Perplexity, Google AI Overviews, Microsoft Copilot and Claude, 50 prompts per engine (44 unbranded plus 6 branded fact checks), one cycle every 72 hours. The snapshot published on 2026-09-18 covers four cycles from 2026-09-08 to 2026-09-17, 1,200 observations in total. /en/report explains in English what a formal delivery contains.

What the dashboard shows

• Answer text: the raw answer for every cycle, engine and prompt, filterable by engine, cycle, prompt and status.
• Sources: the URLs each answer cited, and for each URL which answers, engines and cycles cited it.
• Status vocabulary: how success, failed, blocked, unknown, refused and not triggered are labelled, and which of them enter the denominator. Failed and not-triggered runs are listed as they are.
• Metrics: numerator, denominator and exclusions of the brand mention rate, peer brands measured on the same basis, source types and the most-cited domains.
• Delivery files: each published monthly report (August 2026 so far) as PDF, plus a Markdown bundle (public redacted edition, no screenshots) of that month's prompts, answers and source URLs, each with a SHA-256 checksum. The bundle covers only the prompts behind that report, not the dashboard's cycles.

What is public and what is not

A snapshot updated by cycle, not a showcase

The dashboard is not live data: each update states its publication date and latest cycle at the top of the page. It shows where a young brand actually stands on the buyer questions it chose, including cycles with a very low mention rate, rather than a hand-picked success story. Miaowa GEO does not promise any uplift in mention rate or ranking, and no figure on the page should be read as a performance commitment.

Using it to test vendor claims

1. Check statuses: if any vendor's sample contains no failed or not-triggered records, ask where they went.
2. Check the denominator: confirm that mention rate counts only unbranded prompts that produced a normal answer, and that branded fact checks are handled separately.
3. Check the evidence form: can you read each answer, its source URLs and a screenshot, or only dashboard figures?
4. Check task IDs: can a single observation be located reliably, and do technical retries reuse the original ID?
5. Write these into your acceptance terms; see /en/help/pre-contract-acceptance-checklist.

Can I see a real client's full report?

Not publicly. Client evidence is used only for that client's own observation, verification, reporting and support, and published customer cases require written authorisation. /report uses Miaowa GEO's own brand; authorised customer cases and evaluation records are at /en/cases.

Is the free check result page the same as /report?

No. The free check at /en/geocheck audits website machine-readability and previews 5 suggested buyer prompts; it does not run prompts on AI platforms. /report is a real project dashboard in the form paid monitoring is delivered.

Is there an English version?

The dashboard interface is Chinese only, at /report. The English page /en/report describes in words what a formal delivery contains.

Last updated 2026-09-19. The capabilities page and the pricing page are authoritative where they differ.