PROOF

Proof, not promises.

Six Hong Kong engagements from the Mercury GEO desk. Codenamed for client confidentiality. No invented #1 rankings — every claim maps to a shipped artifact you can inspect.

Premium dining · Hong Kong

Omakase Desk

Target query · best omakase in Hong Kong

Before

AI defaulted to listicles and review platforms. The kitchen was world-class. The website was not citeable.

After

Criteria-led answer architecture, Restaurant schema, bilingual FAQ. Clearer shortlist eligibility — less dependence on third-party roundups.

LocalBusiness + menu JSON-LD EN / 繁中 FAQ cluster 90-day GEO backlog

OutcomeAI shortlist eligibility for comparison queries

Retirement finance · Hong Kong

Annuity Path

Target query · retirement annuity · 退休年金

Before

Jargon-heavy, fragmented pages. Assistants could not extract definitions or next steps. Affiliates owned the answer.

After

Bilingual answer hub: definitions, eligibility, process, decision criteria. Cleaner path from discovery to consultation.

Glossary + FAQ schema FinancialService entity map Compliance-aware copy loop

OutcomeHigher citeability on definitional queries

Telecom · Hong Kong

Premium Mobile

Target query · best premium mobile operator

Before

“Best mobile” collapsed into price tables. Premium positioning flattened. No extractable proof for plans, SLAs, or coverage.

After

Offer matrices AI can quote. Brand entity separated from competitors. Sales-usable proof blocks on-page.

Plan / offer entity structure Premium differentiation matrix Canonical + schema hygiene

OutcomeSeparated from price-led recommendations

Casual-fine dining · Hong Kong

Yakitori House

Target query · best yakitori · 燒鳥串

Before

Menu lived as photos. Models could not name dishes, price band, or booking path.

After

Menu and entity clarity, LocalBusiness schema, bilingual FAQ, criteria-led answer blocks.

Menu entity LocalBusiness schema Bilingual FAQ

OutcomeCiteable for 燒鳥 comparison queries

Pet services · Hong Kong

Pet Farewell

Target query · quality pet funeral services

Before

Sensitive queries surfaced spam directories. Trustworthy operators were invisible to assistants.

After

Step-by-step process pages, trust modules, local service schema, conversion paths that respect the moment.

Process pages Trust modules Local service schema

OutcomeTrustworthy surfacing on high-intent queries

B2B services · Hong Kong

LLM SEO Desk

Target query · LLM SEO provider Hong Kong

Before

Category terms were owned by generic agencies. No technical GEO surface, no audit language models could cite.

After

Technical GEO stack: schema, llms.txt, AI robots, audits. Same operating system used on Mercury’s own site.

llms.txt + crawl policy 6-dimension audit Entity pages

OutcomePage-one + AI Overview citations on category terms

Next

Your category already has an AI answer. Let's make it you.

Start with a 48-hour baseline AI visibility audit — the same first step every engagement above took.