Case study · Online jewelry store, Hong Kong

Nine Carats: the jewelry store no AI tool has named yet

How the audit gave them clear direction on what to fix first and where to focus.

ninecarats.com · Audit Sep 2026 ·

What the audit changed, in short

  • MarketHong Kong first, not North America. Taiwan added.
  • SpendYes to a paid place in the guide AI quotes. No to a sustainability certificate.
  • Biggest issueGoogle listed 2 of 84 pages. Now 36.
See all six decisions ↓

Before the audit

Nine Carats had a beautiful site, a showroom in Central, Hong Kong, and delivery to Hong Kong, Singapore, the US and Canada.

The team planned to grow in North America, which they saw as the bigger market. Taiwan was not on their list. The brand's story led with sustainability.

Sassy Hong Kong, a local lifestyle site, had offered them a paid feature, and they were not sure it was worth the money. They had no Google Business Profile. And when buyers asked an AI tool where to buy lab-grown diamonds, Nine Carats never came up.

Six decisions the research changed

The audit was not only a fix list. Most of its value was in what the team decided to do next, and what they decided not to spend on.

  1. Decision 1

    Hong Kong first, not North America

    Before: North America looked like the bigger prize.

    What the research showed:

    • US: biggest market, but searches are dropping
    • US AI answers already name the same seven big brands
    • Hong Kong: Chinese searches are growing, with light competition
    • The showroom is in Hong Kong

    After: Hong Kong became the first market, in English, then Chinese.

    Change in searches over one year

    −33%

    US, lab-grown engagement rings

    +26%

    Hong Kong, 培育鑽石 (lab-grown diamond in Chinese)

    Hong Kong fixes in the next release.

  2. Decision 2

    Taiwan added to the plan

    Before: Taiwan was not a target market.

    What the research showed:

    • Same search demand as Hong Kong
    • Far less competition
    • Beginner questions more than doubled in a year
    • The brand's designer is based there

    After: Taiwan follows Hong Kong Chinese and reuses most of that work. The Taiwan pages will explain how to buy, as the site does not deliver there yet.

    Monthly searches for 培育鑽石
    • Hong Kong2,400
    • Taiwan2,400

    Planned for month 3.

  3. Decision 3

    The words buyers use, not the ones that sound right

    Before: "Lab grown diamond engagement rings", the US phrase, looked like the one to use everywhere.

    What the research showed:

    • Almost nobody in Hong Kong searches that phrase
    • Buyers type plain "lab grown diamond"

    After: Each country's page uses the words its buyers type.

    Monthly searches in Hong Kong
    • lab grown diamond1,600
    • lab grown diamond engagement ringsabout 20

    Built into the Hong Kong fixes.

  4. Decision 4

    Sustainability dropped as a selling point in search

    Before: Sustainability led the brand's story online.

    What the research showed:

    • Few buyers search for it
    • The brands AI names as sustainable all hold an outside certificate
    • Nine Carats had none

    After: Dropped it for now, and skipped a certificate the demand does not justify.

    Share of lab-grown diamond searches, in every country checked
    • Sustainability searches1.5 to 3%
    • Everything else97% or more

    Decided 8 Oct 2026.

  5. Decision 5

    Yes to the paid Sassy Hong Kong placement

    Before: Sassy had offered a paid feature. The team was not sure it was worth it.

    What the research showed:

    • Perplexity quoted Sassy's buying guide in every Hong Kong answer
    • Nine Carats was not in it
    • Guides like this shape AI's "where to buy" answers more than a brand's own site

    After: Went ahead with the paid placement in the guide AI already quotes.

    Perplexity answers that quoted Sassy Hong Kong's guide
    3 of 3 tries, before and after the fixes

    Publishing November 2026.

  6. Decision 6

    A Google Business Profile

    Before: There was none.

    What the research showed:

    • Hong Kong competitors had hundreds of Google reviews each
    • Nine Carats had no profile

    After: The team is setting one up.

    Google reviews, Hong Kong (4 Nov 2026)
    • DIABOND1,100
    • Novita300
    • Nine CaratsNo profile

    In progress.

Search numbers are Google data from September 2026 and are approximate. AI answers: ChatGPT, Perplexity and Claude, 3 tries per country.

What has changed so far

Fixes went live on 21 Oct 2026. The first check was 14 days later, and Google is still catching up.

36

of the site's 84 pages now show up in Google

Up from 2 before the fixes (18 Nov 2026).

60

out of 100: how easy the site is for AI to read

Up from 34, using my audit checklist.

2 of 3

AI tools that describe Nine Carats correctly

Perplexity still mixes it up with a different company in Australia. Same as before the fixes.

0

AI answers out of 45 that recommend Nine Carats

Not yet. Same as before the fixes. More on why below.

Pages that show up in Google, out of 84 on the site
  • 9 Oct, before the fixes2
  • 4 Nov19
  • 9 Nov35
  • 18 Nov, homepage now in Google36

Having a page on your site does not mean Google shows it. Google decides which pages to read and list in its results, and a new or small site can wait weeks. A page that is not listed cannot be found in Google search or in Google's AI answers, however good it is. Before the fixes, Google listed only 2 of the 84 pages: almost every product was invisible. The number climbs as Google gets to each page, which is why it keeps rising after the fixes went live.

Timeline

  1. 28 SepAudit delivered
  2. 9 OctStarting point recorded
  3. 21 OctFixes go live
  4. 4 NovFirst check
  5. 18 NovHomepage in Google
  6. DecNext check

What has not moved yet, and why

AI tools do not recommend Nine Carats yet: 0 of 45 answers. AI search is three separate jobs, and each one moves at its own pace.

  1. 1 · Research

    What people search

    It set which market comes first, which words to use, and what to spend on.

    Done in the audit.

    Done

  2. 2 · Your own site

    What your pages say

    The audit listed what stopped Google and AI from reading the site. 9 fixes went live on 21 Oct, and Google now lists 36 of 84 pages. Readable is not yet recommended.

    The audit's fix list. Their engineer built it, and I checked each fix live.

    Moving

  3. 3 · Other people's pages

    What others say about you

    AI recommendations come from here. The audit showed which guide AI quotes in Hong Kong, and that Nine Carats was not in it. That led to the paid placement.

    Mapped in the audit. Their own team does the outreach.

    Not moved yet

What this case study covers

The client chose to build the fixes with their own engineer, so this case study is about the audit and the tracking.

  1. AuditIn this case studyResearch and site check, delivered 28 Sep.
  2. FixesBuilt by the client's engineer from my specs. I checked each one live.
  3. TrackingIn this case studyA check every few weeks since 9 Oct, measured against the starting point.

What happens next

The next check is in early December 2026: same questions, same AI tools. I will update this page with the results.

The Sassy Hong Kong placement should be live by then. If Hong Kong AI answers start naming Nine Carats, I will credit the placement first, not the site fixes, because AI tools were already quoting that guide.

Next on the site: the Hong Kong fixes. Each market gets its own set, built around what buyers there search for and ask. Hong Kong comes first, then Hong Kong Chinese, then Taiwan.

Last updated 20 Nov 2026.

Try this on your own store

  1. Ask ChatGPT "Where can I buy [what you sell] in [your city]?". Is your store named? Who is?
  2. Ask "What is [your brand]?". Does it get the right company, country and products?
  3. Search Google for site:yourstore.com. If it shows far fewer pages than you have products, Google is not seeing your store.
  4. Read your delivery terms on the FAQ, a product page and checkout. Do they say the same thing?
  5. When an AI tool names your competitors, look at which guide or blog it quotes. Are you in it?

Behind the work

The detail, for anyone who wants it: what I did, what was holding the site back, what was fixed, and what I got wrong.

What I did

  • Looked up what buyers search for in six countries and two languages.
  • Asked ChatGPT, Perplexity and Claude the questions buyers ask, three times each per country, and noted which brands they named and which websites they quoted.
  • Checked the site the way Google and AI tools read it, and wrote a list of fixes.
  • Recorded where everything stood before any fix went live, so we could measure the change.

The client's own engineer made the fixes on 21 Oct 2026. I wrote what to change, checked every fix on the live site, and handled the Google steps.

What was holding the site back

  1. Problem 1

    Google could see 2 of 84 pages

    The site sent Google to the wrong version of its web address. Google kept hitting a dead end, so almost every product page was missing from search.

  2. Problem 2

    One AI tool confused the brand

    Perplexity mixed Nine Carats up with a different company with a similar name, in Australia. ChatGPT and Claude got it right.

  3. Problem 3

    AI could not read the FAQ answers

    Shoppers saw each answer when they clicked a question. AI tools reading the page only saw the questions.

  4. Problem 4

    Four different answers on shipping

    The FAQ, a banner and the product pages each said something different about delivery, and checkout said a fourth thing. An AI tool could quote any one of them to a buyer. This one is still being fixed.

What was fixed

  • Google is now sent to the right web address, so it can find every page.
  • The FAQ answers are now readable by AI tools. Shoppers still click a question to open its answer, as before.
  • A short file now tells AI tools who the brand is and what it sells.
  • Every product now carries clear details AI tools can read: sizes, price, returns and delivery.
  • The diamond grading lab, IGI, is now named on every product page.
  • The company name, showroom and workshop addresses are now in a form search engines read.
  • Page titles now use the words buyers search.

What I got wrong

The biggest lesson: a testing tool's reading of your site is not your site. Three of my mistakes reached the client, and my audit missed the biggest real problem.

  1. Mistake 1

    A layout problem that wasn't there

    One testing tool reported a layout problem, and I put a fix for it in the audit. A second tool showed the pages were fine. I had a fix drafted before I checked, and I had to send the client a correction.

  2. Mistake 2

    A product label Google does not accept

    I recommended marking products "made to order" using the general web standard, without checking Google's own rules. It went live and Google rejected it on every product. Fixed within a day.

  3. Mistake 3

    A fix I called checked too early

    After launch, I checked the homepage pointed Google to the right website, but not the full address. It pointed to a duplicate copy. That took another fix and four weeks before Google listed the homepage.

  4. Mistake 4

    The biggest problem, missed

    My testing tools said Google could read the site. Google's own dashboard showed it had listed 2 of 84 pages. I only saw that after the audit was delivered.

What I check now

  • Every finding is confirmed by a second tool before it reaches you.
  • Google's own count of your pages comes before any score.
  • Product details are checked against Google's rules, not only the general standard.
  • After a fix goes live, I check the exact address, not just that something is there.

Request a free store check

Send your brand name and work email. I'll reply by email, usually within 2 days, with what AI says about your store today and one fix you can make yourself. No obligation: nothing else starts unless you ask.

A custom assessment of how AI tools describe and recommend your products.

Prefer email? Email me your details