This page carries the projects we took on, anonymised baseline audits from our free checker, and the assessments where we decided a site was not suitable and declined. Every published figure comes with its sample size, platform, observation window and locale, and every conclusion states what it does not extend to.
Each one follows the same skeleton: baseline, what was done, a like-for-like re-test, what we cannot claim credit for, limits, and the client's authorisation.
The product and marketing teams at Miaowa GEO and Zhixue Classroom completed a roughly 30-minute video discussion covering Miaowa GEO's capabilities, brand and product demonstration. Both sides confirmed collaboration intent and will continue defining scope and implementation arrangements.
Read in fullAfter one phase of GEO optimisation, SupplyICs began entering supplier candidate lists in unbranded procurement questions on several AI products. Organic traffic subsequently increased, and an email connection ultimately led to a completed sale. This case discloses the diagnosis, remediation method, selected AI-answer samples and attribution limits.
Read in fullSailin is a real solar bird-guard-mesh manufacturer in Anping, Hebei: a complete product system, OEM capability, pages written in the buyer's own language — and Perplexity and Google AI Overview already cite it. Yet in unbranded overseas-buyer prompts it appears in 4 of 14 answers while the category leader appears in 10 of 14. The gap is not the product but three pieces of unpublished evidence. This is Miaowa GEO's first client-authorised public case: how the baseline was measured, where the gaps are, and how 90 days of remediation will close them — methodology fully disclosed.
Read in fullIn August 2026 five export-oriented businesses submitted their websites to our free checker. Their baseline audits failed in different places — every one invisible to the owner: firewalls mis-killing AI crawlers wholesale, template robots.txt rules barring them, Chinese-English twin sites that never declare each other, structures machines cannot read. This page publishes five anonymised slices and the export-subset methodology. None of the subjects is our client; with no authorisation, none is named.
Read in fullEngagements we turned down, and the checkable findings behind each call. We publish them not to judge anyone, but because the criteria are ones you can apply to your own site.
A cross-border storefront we turned down, and the four checkable findings behind that call.
Read in fullA B2B-marketplace storefront we turned down: five findings checkable on its own profile page, three sets of public data on how answer engines verify, and why we would rather lose the deal — GEO on that foundation hurts overseas buyers first, and the answer ecosystem in the end.
Read in fullA cross-border accessories storefront we declined. It had hundreds of products, multilingual content, enquiry routes and card checkout, and a real-company trail could be found. Yet the brand operator, legal seller, product evidence, payment, fulfilment and after-sales responsibility did not form one consistent public fact chain. More importantly, the operating team could not promptly control and correct the site, and would not complete factual remediation first. We declined to increase GEO visibility before fact governance was in place.
Read in full