ChatGPT Search
Synthesises live web and other available sources in conversation form. Citations shift with the question, time and search results. Useful for watching whether a brand enters comparison and recommendation answers.
AI Search
What is AI Search?
AI Search retrieves, compares and synthesises multiple sources before answering directly. Users may finish early research in ChatGPT Search, Perplexity, Gemini or Google AI Overview—without browsing ten blue links one by one.
Search behaviour is changing
From “finding links” to “getting an answer”
Traditional search mainly ranks candidate pages. AI Search interprets the question, retrieves material, verifies sources, then generates a synthesised answer.
This shift does not mean Google or SEO disappear. Generative answers still lean heavily on crawlable pages, search indexes, knowledge graphs and third-party sources. The difference is that brands must compete for rankings, for correct model understanding, and for a place as a citable source inside the answer.
For Hong Kong brands, buyers often ask AI—“which firm fits,” “what risks to watch,” “what drives price”—before WhatsApp or a meeting. If your brand is missing from the answer, or if service area and credentials are wrong, a shortlist may form before you ever speak to the customer. The usual next step is a visibility diagnosis: see the AI Visibility Audit.
Versus the classic SERP
Ten blue links and generative answers solve problems differently
SEO teams already optimise titles, snippets and rank position. AI Search also asks whether an answer is extractable, whether the brand is verifiable, and whether sources corroborate one another.
| Dimension | Classic SERP | AI Search answer |
|---|---|---|
| User work | Browse many results and compare manually | Read a synthesised answer first, then open sources as needed |
| Brand ticket in | Strong rank and a compelling snippet | Selected into the generated answer and described correctly |
| Content unit | Whole-page topic and keyword coverage | Definitions, comparisons, steps and limits that extract cleanly |
| External signals | Links and page authority | Entity consistency, credible third parties, cross-platform corroboration |
| Failure mode | Not ranking / low CTR | Absent, misdescribed, or only endorsed by sources you do not control |
| Attribution difficulty | GSC / Analytics relatively clear | Zero-click and incomplete referrers are common |
So “AI Search optimisation” is not a project unrelated to SEO. The site still needs to be indexable, fast and well structured. The difference is that high-intent questions must be written answer-first, with separate mention / citation / accuracy metrics. In practice the two run together: SEO owns indexation and topical authority; GEO / AIO own answer visibility and description accuracy. Cross-check the Hong Kong GEO guide and the AIO method.
The same query can surface different sources in classic results versus AI Overview / chat search. Do not assume “rank one means citation,” and do not assume “citation means stable clicks.”
Major answer surfaces
AI Search platforms do not use sources the same way
Perplexity
Source links are relatively explicit—suitable for logging which pages and passages are treated as answer evidence, and for building citation URL records.
Google AI Overview / AI Mode
Combines Google’s search index, pages, video and other results. SEO foundations still matter; local intent and business profile data are often used together.
Gemini
Closely tied to Google’s ecosystem and multimodal content, but sources still vary by query. Sample separately—do not equate it with AI Overview.
Microsoft Copilot
Combines search with work workflows. Brands need clear, verifiable, machine-readable content—especially B2B explainers and documentation pages.
Claude and other entry points
Live search capability, citation UI and regional availability keep changing. Sample against your target audience; avoid measuring only one interface.
See the full breakdown of platforms and common citation sources →
Impact for Hong Kong brands
Local buyers mix Traditional Chinese, English terms and brand names
The same person may search “香港 GEO 服務,” type “geo service hk,” then chase ChatGPT with an English product name. Content and entities need cross-language consistency—without replacing natural answers with keyword stuffing.
Comparison queries
Which company, which plan, how A differs from B—usually close to a vendor shortlist.
Explainer queries
What GEO is, how AI Search works—good opportunities to build topical authority.
Local-intent queries
Hong Kong, pricing factors, process, compliance, language and service area must be clearly verifiable in content.
Teams often underestimate mixed-language recall. A page written only in English may answer “what is GEO” well for expatriate buyers, yet still leave Traditional Chinese “GEO 是甚麼” reliant on competitors or generic sources. Likewise, a Traditional Chinese brochure page stuffed with slogans may never give the model a clean extractable answer to “GEO agency Hong Kong timeline.” The practical fix is not dual keyword stuffing—it is one consistent set of facts rendered as natural answers in both languages, with internal links that keep the topic connected.
Hong Kong buyer journey
AI Search usually enters at understand, compare and verify
Not every Hong Kong buyer places an order through AI. But more people form an early view from generative answers before contacting sales.
Understand the problem
Users ask “what is it / is it for SMEs / what are the risks.” Without definition pages and stated limits, answers may cite competitors or generic encyclopaedic content.
Form a shortlist
Comparison questions place several vendors in one answer. Being correctly classified and described matters more than winning the longest blurb.
Verify credentials and local conditions
Users chase service area, language, pricing factors, professional credentials or timeline. Directories, author pages and business profiles start shaping credibility.
Move to the site or a conversation
Some users click sources; others take the AI summary straight to WhatsApp. The official site still needs to receive high-intent visits and clear up misstatements with consistent facts.
In practice: education content serves understanding; service and comparison pages serve shortlisting; FAQ and evidence modules serve verification; audits and fixed retests serve internal decisions. The Hong Kong page GEO for Hong Kong businesses covers bilingual and local source setup; writing specs live in guides.
Measuring AI Search
Watch exposure, citation and conversion separately
AI answers may build brand awareness without clicks—or send a small number of high-intent referrals. Organic traffic alone is an incomplete judgement.
| Layer | Question | How to log it |
|---|---|---|
| Mention | Did the brand enter a relevant answer? | Mention rate and context on a fixed query set |
| Citation | Which page or external source was listed? | Citation URL, position, platform and date |
| Accuracy | Are services, location and limits described correctly? | Check predefined fact fields one by one |
| Action | Did the user visit, enquire or become a qualified lead? | GA4 referral, forms, WhatsApp and CRM source |
Platforms may not pass a full referrer, and zero-click exposure never appears in GA4. Use site analytics and fixed query sampling together.
What brands should measure
Build retestable operating metrics before growth storytelling
Leadership often asks whether “doing AI Search worked.” The answerable version is not “guaranteed to rank.” It is: which buyer questions we were absent from, where descriptions were wrong, which controllable signals we fixed this quarter, and how trends moved on retest. At minimum, lock the following operating metrics:
Query coverage
Of high commercial-value queries, how many already have a unique answer page or service page covering them.
Mention / citation rate
Under the same platforms and logging rules, the share of times the brand is mentioned or linked to the official site.
Entity accuracy
Whether name, location, services, pricing factors and limits are restated correctly.
Source mix
Whether answers lean on the official site, professional directories, media or UGC—and what a healthy mix looks like for the category.
Assisted enquiry quality
Whether referral / recommendation enquiries are better aligned to service scope—not only overall sessions.
Governance cadence
Whether content, Schema, business profiles and directories have owners and review dates.
Do not compare opportunistic screenshots across platforms, phrasings or months and declare a “win.” A sounder approach is quarterly retesting of the same query set, separating site changes you control from platform-side shifts. Live market examples live in GEO / AI Search case analyses; service options live in Hong Kong GEO services.
Any mention or citation rate is a sampling observation—shaped by model updates, indexing and competition—and cannot be extrapolated into a long-term guarantee.
If leadership needs a communication frame: report controllable work (pages shipped, entity fixes, source repairs) separately from observed AI outcomes (mention / citation / accuracy trends). That separation prevents the team from promising what platforms cannot guarantee—while still showing progress the business can audit.
For practical next steps in Hong Kong, most teams start by locking a bilingual query set, running a baseline Audit, fixing entity contradictions, then rewriting the highest-intent service and comparison pages into answer-first structures. Platform choices and source mix can wait until that foundation is retested.
Want to know whether your brand currently appears in AI Search answers—and whether it is described correctly?
Request an AI Visibility AuditFAQ about What Is AI Search? A Guide for Hong Kong Brands
Will AI Search replace Google?
Not wholesale today. AI answers may absorb some explanatory, comparison, and research queries, while classic search, Maps, shopping, and site browsing remain important.
How is AI Search different from a blue-link SERP?
Classic search ranks links. AI Search retrieves and synthesizes sources into an answer and may show only a few citations—so brands compete for both rankings and answer visibility.
Which AI platforms should Hong Kong brands sample?
Sample ChatGPT Search, Perplexity, Gemini, Google AI Overview / AI Mode, Copilot, and other relevant surfaces by audience—and log Traditional Chinese, English, and region settings separately.
If we rank on Google, will we appear in AI Search?
Not automatically. Rankings can aid discovery, but AI also weighs extractability, entity clarity, source corroboration, and what is retrieved in that session.
How should brands measure AI Search performance?
Use fixed queries, platforms, regions, and fields to track mentions, citations, description accuracy, and referral traffic. One screenshot is not a trend.
Want a baseline of how AI answers describe your brand?
Request a free Audit