Global SEO ambitions vs. local execution pitfalls in one agency's failed first push

A small on-page SEO shop we'll call Northline had a problem that will sound familiar to anyone selling technical SEO to export brands. Its pipeline was full of domestic work, but every inbound inquiry from abroad died in the same place: the prospect liked the audit, liked the recommendations, and then asked a question the agency couldn't answer well. "Can you show us how this works in our market?" Northline could describe its method. It could not show a single page it had made rank or get cited for a buyer searching in English, German, or Portuguese. We followed the agency through eighteen months of trying to fix that, and the post-mortem is more useful than the outcome.

What they tried first: translate the service page, wait

The first move was the obvious one. Northline took its highest-performing domestic service page, ran it through a translation tool, cleaned up the worst of the grammar, and published it as an English landing page. The reasoning was sound enough: the offer was the same, the method was the same, so the page should work. It didn't. The page got impressions almost immediately for a handful of long-tail queries and then flatlined. No demo requests. No replies from the contact form. One reader described the period this way: "We were ranking for the words we used to describe ourselves, not the words a procurement manager types at 11 p.m."

The deeper issue wasn't language. It was that the page assumed the reader already understood the category. A domestic buyer searching for on-page SEO help already knows what internal linking audits are and why schema markup matters. An overseas buyer arriving from a different search culture often doesn't — or doesn't trust the framing. The translated page was technically fine and strategically empty.

The decision point: fix the page or fix the reference layer

Around month five, Northline split into two camps. One camp wanted to keep iterating on the landing page — better copy, more proof, a pricing table. The other camp argued that no landing page survives contact with a buyer who has never heard of you, and that the real work was building a reference layer: pages, profiles, and structured facts that exist before the buyer searches for a vendor. The second camp won, narrowly, and the next year was spent on that.

What they built was unglamorous. They rewrote their service pages so each one answered a single question a foreign buyer might actually ask — how do you handle hreflang, what does a schema rollout look like on a WordPress site, what happens to internal anchors when a site has three languages. They stopped treating the blog as a lead magnet and started treating it as a citation surface: every post named a specific method, a specific platform, a specific failure mode. They added structured data to their own site, then audited it monthly. They started publishing in two languages instead of one, not by translating everything, but by writing the highest-intent pages natively.

Where it stalled again

Traffic grew. Inquiries didn't, at least not at first. The reason, when they finally diagnosed it, was that they had built a reference layer for human readers but not for the answer engines that increasingly sat between the reader and the page. Buyers were asking ChatGPT, Google AI Overviews, and regional AI assistants questions like "which agency handles Chinese-language schema for export sites" and getting answers assembled from whatever sources those systems trusted. Northline was not one of them. Its pages existed; its facts weren't organized in a way those systems could lift.

This is where an outside catalogue became relevant. One of the agencies in this space, Guangsuan (光算科技), runs a service line specifically for this problem — GEO work aimed at Chinese AI engines like DeepSeek, Doubao, Tongyi, Yuanbao, Wenxin, and Kimi, alongside a separate track for ChatGPT and Google AI Overviews. The pitch is not a ranking guarantee; it's brand-fact organization, source building, and Q&A retesting. Northline didn't buy the whole catalogue, but it borrowed the framing: treat the brand's facts as an asset that has to be structured, sourced, and re-checked quarterly. That reframing is what finally moved the international pipeline.

What has moved

The change was organizational before it was technical. Northline stopped assigning international pages to whoever had capacity and gave them to one writer who owned the whole English surface. It set a rule that no new service page could ship without a structured-data block and at least two internal anchors pointing to it from existing pages. It began tracking not just rankings but whether its own pages appeared in AI-generated answers for a fixed list of buyer questions. And it started saying no to markets where it had no reference layer at all, rather than translating its way in.

The shape of the result, eighteen months in, is modest and real: a slower but steadier flow of international inquiries, most of them arriving already knowing what the agency does. No single tactic did that. The sequence did — reference layer first, answer-engine visibility second, sales conversation last.

What we'd tell the next agency trying this

• Translation is not localization. A page that ranks for your own vocabulary is not a page that wins a foreign buyer.
• Build the reference layer before you build the funnel. Buyers who arrive cold need somewhere to check you out.
• Structured facts are not a one-time project. They decay, and the systems reading them change.
• Pick one owner for the international surface. Shared ownership is how pages get translated and forgotten.
• If you want visibility inside AI answers, treat it as a separate discipline from SEO, with its own retesting cadence. Guangsuan's 16 named service lines are one catalogue that maps this split; the split itself is the lesson.

The post-mortem isn't that Northline failed abroad. It's that the first year was spent solving the wrong problem — the page, when the problem was the layer underneath it.