AI Growth Hub
← AI SEO

SEO for a New Trilingual Site: What We Did in the First Month (and What We're Still Measuring)

technical-seo ai-search-visibility geo google-search-console

The short version

In its first month, a new site’s biggest SEO problem isn’t ranking. It’s whether search engines come and crawl it at all. Google started showing our pages within the first few weeks. Bing read our sitemap and, two days later, still hadn’t crawled the homepage.

This is a log of what we did, what we saw, and what we are still waiting on. The parts with no result yet are marked that way. We’ll update this post once the numbers are in.

The site in question

The site is BC Class 5&7, a practice site for the British Columbia driver knowledge test. It has 530+ practice questions, and every page exists in English, Simplified Chinese and Traditional Chinese. There’s no account or sign-up, and every question on the website is free to answer.

This is our own project, built by the same people who write vozai.net. We’re writing it up because it’s a clean example of a problem many cross-border sellers hit when they launch a second-language store: a brand-new domain (in this case a subdomain), no backlinks, and several language versions of every page. The topic is driving tests rather than products, but the crawling and indexing side works the same way.

The technical baseline we got right on day one

None of this is advanced. It’s the checklist we’d hand anyone launching a multilingual site, and every item below was live when we checked on October 4:

  • Every page has a canonical tag pointing to itself. The English page doesn’t claim to be the original of the Chinese one.
  • The three language versions reference each other with hreflang, plus an x-default.
  • One sitemap lists every page: 1,800+ URLs across the three languages. robots.txt allows crawling and gives the sitemap address.
  • Inner pages carry BreadcrumbList structured data.
  • Pages are static HTML, served compressed through Cloudflare.

If you’re running a zh/en store, two common mistakes are hreflang tags that don’t point back to each other (page A lists page B, but B doesn’t list A) and canonical tags that point every language version at the English page. The second one tells Google your Chinese pages are copies of the English one, so Google may index only the English version.

What the first month of data looked like

Google Search Console started reporting impressions within the first few weeks, followed by a handful of clicks. Volume was small, as you’d expect for a new subdomain with no links.

The more useful finding was where the impressions came from. Most of them were long-tail searches that matched the exact wording of a practice question, things like “soft shoulder sign meaning”. For a store, the equivalent is that early impressions are likely to come from very specific product-attribute searches, so those pages deserve your attention first.

One trap in Search Console: the Pages (indexing) report can lag well behind the Performance report. When we read the data on October 4, Performance ran through October 2, but the Pages report was dated September 20. If you use the Pages report as “what’s indexed right now”, you’re looking at a picture two weeks old.

Bing: sitemap read, nothing crawled

Bing was the slow one. In Bing Webmaster Tools, the sitemap showed as read with about 1.6K URLs discovered. Site Explorer still said “No data available”, and URL inspection on the homepage said “Discovered but not crawled”.

This matters more than Bing’s share of web search suggests. Microsoft’s documentation for Microsoft 365 Copilot says that when web search is on, Copilot may fetch information from the Bing search service to ground its answers. Our inference: a site Bing hasn’t crawled is unlikely to turn up in those answers.

We are adding IndexNow to our release process. IndexNow is the protocol that lets a site tell participating search engines a URL has changed instead of waiting for a crawl. Two details from the official sources shaped how we’re setting it up:

  • Bing’s IndexNow getting-started page says “you should publish only URLs changing (added, updated, or deleted) since the time you start to use IndexNow.” So we’re not pushing all 1,800+ URLs in one go, only the ones that change.
  • The list of participating search engines has Bing, Yandex, Seznam, Naver, Yep, the Internet Archive and Amazonbot. Google isn’t on it. IndexNow does nothing for your Google indexing.

Before setting up IndexNow, check that your CDN isn’t blocking the crawler in the first place. If you’re on Cloudflare, it changed its AI-bot defaults on September 15 (we covered what the new defaults block), and bot or firewall rules can also catch ordinary search crawlers. A ping to Bing is pointless if Bingbot gets a challenge page when it arrives.

Status: in progress, not live yet. Results to follow.

Why Google showed a globe instead of our icon

In Google’s results, the site appeared with the default globe icon, and the site name showed as the root domain instead of the site’s own name.

The most likely cause was simple. The site’s favicon was an SVG inlined as a data: URL. Google’s favicon guidelines list the supported formats as “BMP, GIF, ICO, PNG, JPEG, PPM, and TIFF”. SVG isn’t on the list. Browsers render an inline SVG favicon without complaint, so you’d never notice the problem by looking at your own site.

The site name is a separate setting. Google’s site name documentation says a site is defined by its domain or subdomain, so a subdomain can have its own name. It also says that if a subdomain’s homepage has no structured data, the domain-level site name may be used as a fallback. That’s most likely what happened to us.

The fix we’re putting in:

  1. Add real icon files: a 192×192 PNG, an ICO, and an apple-touch-icon.
  2. Add WebSite structured data to the homepage with name and alternateName.

Status: planned, not yet visible. Google says recrawling for favicons “can take anywhere from several days to several weeks”.

With zero backlinks, the question was where a legitimate first link could come from.

The most promising places were B.C. pages that already list driver knowledge test resources: public libraries, a literacy organization and a traffic-safety site. These pages exist to point people to outside practice material, so a free practice site is something they might actually want to list. We sent requests to four of these pages. No replies yet.

We also ran into a paid option. The traffic-safety site on that list also openly sells “Contextual Link Insertion” in its existing articles. We didn’t buy it. We sent it the same free listing request as the others; if it replies with a price, we’ll decline. Google’s spam policies say paid links are fine “as long as they are qualified with a rel=“nofollow” or rel=“sponsored” attribute”. Google’s page on qualifying outbound links says links with these attributes “will generally not be followed”. A properly labeled paid link is unlikely to help rankings, and an unlabeled one breaks the policy.

Community platforms taught us something the hard way. An old but barely used, low-karma Reddit account posting a link to its own site had the posts removed by the platform’s filters, and a community tool showed the account’s Contributor Quality Score at the lowest tier. As far as we can tell, the filter judged the account, not the content: an account with almost no history whose first posts link to one site looks like self-promotion. If you plan to use communities, the account needs a history of ordinary participation first.

This post is a link too, and we kept it within the same rules: it says plainly that the site is ours, and the site’s blog is the only other place we link.

What we chose not to do

We didn’t add FAQ structured data. In August 2023 Google said that FAQ rich results “will only be shown for well-known, authoritative government and health websites.” A practice site isn’t one of those, so the markup would add work and change nothing in search results.

We didn’t build a page for every city. It’s tempting to generate “driving test practice in Richmond”, “in Surrey”, “in Burnaby” and so on. Our questions apply to all of B.C. and nothing in them is city-specific, so a page per city would only repeat the same content.

And we didn’t stuff keywords into titles. Each title says what the page is. Nothing like “BC Driving Test Practice Free ICBC Knowledge Test 2026”.

What’s next

We’ll reread the data around October 18 and add an update to this post covering three things: whether Bing has crawled the homepage, whether the favicon and site name changed in Google, and whether any of the four pages we wrote to added a link.

阅读本文中文版: 三语新网站的第一个月 SEO：我们做了什么，还有哪些在等结果

Related Articles

AI Traffic Control Compared 2026: How Cloudflare, Akamai and Vercel Classify Bots Differently

Cloudflare split AI traffic into Search, Agent and Training in July 2026. Akamai sorts the same traffic into training crawlers, search crawlers and fetchers. Vercel's AI bots managed ruleset covers training, search and user-generated fetches. Three vendors, nearly identical splits, different names and different defaults. A selection guide with a comparison table, an honest read on robots.txt, and a recommended setup for three kinds of store.

ai-search-visibility technical-seo tool-comparison