Case studies — Mercury GAIO practice
Updated August 2026
Proof, not promises.
Hong Kong GEO work from the last 12 months. Each case shows the baseline, the intervention, and what AI engines could cite after — PCCW, Sun Life, omakase / yakitori, and pet funeral services.
Methodology & anonymization
How these outcomes are measured
Every outcome on this page is measured by Mercury Orbit, our recurring citation-monitoring loop. Weekly prompt panels — frozen at baseline so trends stay comparable — are executed against ChatGPT, Gemini, Perplexity and Google AI Overviews; raw answers are stored server-side so any reported figure can be replayed and audited.
Citation Frequency, Share of Voice and Recommendation Rank are computed per engine and per language (EN / 繁中 / Japanese where applicable), never blended. Deltas are reported against the week-1 baseline established before any intervention shipped.
Named clients on this page (PCCW, Sun Life, QQS hospitality venues, pet funeral services) are the operators who ran the work with us. Campaign internals and unpublished figures stay under NDA. Outcomes below are eligibility and system improvements — not unverifiable “#1” rank claims.
Named operators · bilingual EN / 繁中 · Orbit measurement loop · NDA on unpublished figures
The library
Four Hong Kong engagements, one measurement loop
CASE 01
Telecom · Premium mobile
PCCW · Bruce Lam · HK / GBA
Baseline: Premium-operator consideration answers defaulted to price tables. Brand entity mixed with MVNO / budget operators. Gemini visibility in Hong Kong did not appear on mainland assistants.
- query
- best premium mobile operator
- eligibility
- Stronger framing in premium-operator consideration answers
- separation
- Clearer split from pure price-led recommendations
- cross_border
- Dual-track so GBA citation is not assumed from HK Gemini visibility
Read the full case →
CASE 02
Insurance · Retirement
Sun Life · Adley Low · Inchcape programmes · YMYL
Baseline: Invisible to AI assistants in Hong Kong on category queries. Thin, jargon-led pages; no bilingual extractable definition of 退休年金.
- query
- 退休年金
- citation
- Within 90 days ChatGPT started citing Sun Life alongside AIA and Manulife
- journey
- Cleaner path from AI/search discovery to human advisory conversion
- authority
- Stronger topical authority vs thin affiliate explainers
Read the full case →
CASE 03
F&B · Omakase & Yakitori
Terence Tsang · QQS · 燒鳥串
Baseline: AI shortlists cited third-party lists. No Restaurant/LocalBusiness schema, no bilingual “what defines great omakase / 燒鳥串” blocks, weak entity disambiguation vs similarly named venues.
- queries
- best omakase in Hong Kong · best yakitori · best 燒鳥串
- eligibility
- Clearer eligibility in AI shortlists for premium dining intent
- density
- Stronger on-site answer density for comparison queries
- citation_path
- Reduced reliance on third-party listicles as the only citeable source
Read the full case →
CASE 04
Pet services
Calvin Yau · quality-led · sensitive category
Baseline: Directory-style answers. No step-by-step service process, no local service schema, no extractable definition of “quality” in this category.
- query
- quality pet funeral services
- trust
- Quality-led queries have a citeable process instead of a directory row
- schema
- Local service markup so engines can describe the operator
- path
- Calm conversion from answer → human care, not aggressive CTAs
Read the full case →
Your baseline next
Get the same three metrics for your brand.
The free AI-visibility audit establishes your Citation Frequency, Share of Voice and Recommendation Rank baseline — the same starting point every case above began with.
Written by James Huang, Founder & CEO, Mercury Technology Solutions · Reviewed by the Mercury GAIO Practice · Updated August 2026