Article
International SEO is not translation
International SEO means rebuilding each page for its market, not translating it and adding hreflang. Search intent, formats and authority differ by language.
In Sweden, you might need reindeer bags to sell at all.
That isn’t a keyword, and no list built from your home-market terms would have found it. It’s the kind of fact you only get by looking at the market itself.
I put the rule this way: AI content for international SEO is not translation. You rebuild the piece per market, because search intent and format differ by language.
Why translating a page is not entering a market
A translated page answers the question your home market asked. The market you’re entering may be asking a different one.
Translation carries the original’s assumptions across the border: what the searcher wants, what the page should look like, how the product is described, who vouches for you. Change the language and every one of those assumptions stays exactly where it was.
That’s why international SEO starts with market understanding, not keyword lists. A keyword list tells you what people type. It doesn’t tell you why, or what they expect to find.
What differs by market
Four things change from one market to the next, and a translation touches none of them.
Search intent. The same need can sit behind different queries, and the same query can mean different things. Hypothetically: a query that returns category pages in one country might return comparison articles or marketplace listings in another. You can’t know until you read the results page.
The winning format. Whatever ranks on that page tells you the format Google rewards for that query in that market: a guide, a listicle, a category page, a video. If your page is a different format, a better translation won’t fix it.
How the product is described. Products carry different names, sizes, use cases and objections in different places. That’s the Sweden point. The words are the smaller part of the gap.
Which sources count. Google lists local signals as part of how it works out a site’s intended audience: local addresses and phone numbers, local language and currency, and links from other local sites (Google Search Central, multi-regional and multilingual sites, last updated 10 December 2025). Your home-market links don’t transfer. The publications that matter in one country may be unknown in the next.
What hreflang does and doesn’t do
Hreflang is an annotation that tells search engines which language or region version of a page belongs to which audience. Google says doing so helps it point users to the most appropriate version of your page by language or region (Google Search Central, localised versions of your page, last updated 21 September 2026).
The same documentation sets limits. Google says it doesn’t use hreflang or the HTML lang attribute to detect the language of a page; it uses its own algorithms. And if two pages don’t point to each other, the tags are ignored.
So hreflang is routing. It says which version goes where. It says nothing about whether the version deserves to be there. Treat it as the last step of the work, not the work.
Why each language needs its own content, links and checks
Each language is a separate piece of work in three places: the content, the links and the checks. That’s where the resource barrier appears.
The content is rebuilt for the market’s intent and format. The links come from sources that count locally, not from your home market. The checks confirm the result: that the right version appears, the page matches what ranks, and nothing broke in the markup on the way.
A giant can staff this with people in every country. A challenger can’t buy that headcount. I’ve spent fifteen years running search operations, including COO roles at The Search Initiative, Search Logistics and SUSO Digital, and the mechanism is simple: when the work scales with people, cost grows with every market you add. The hire-or-build decision is the same problem seen from the agency side.
What the work looks like as a process
The process has five steps, and the order matters.
- Research the market’s results pages. Read what ranks for the queries that matter, in the local language, from the local market.
- Decide the format. Choose what the results page rewards, not what you already have.
- Write for the market. Build the piece from that brief, with the local product description and local terms.
- Build local relevance. Earn mentions and links from sources that count in that market.
- Check. Compare the finished page against the results page, then confirm the technical layer, hreflang included.
Step three is where translation gets mistaken for the whole job. It’s one step of five.
Translation versus rebuilding per market
The two approaches differ at every stage, not only in how the text is produced.
| Translation approach | Rebuild per market | |
|---|---|---|
| Starting point | The home-market page | The market’s results pages |
| Search intent | Assumed to match the original | Checked against what ranks locally |
| Format | Kept from the original | Chosen to match what wins |
| Authority and links | Home-market sources | Sources that count in that market |
Where AI-native infrastructure takes over
This section is self-promotion: NON.agency is my company, and the claim below is ours, not an independent finding.
NON.agency’s claim is that a market entry takes a few weeks instead of a six-month programme. The mechanism is the split between judgement and execution. Specialists decide the format and judge whether a piece fits the market. AI-native infrastructure runs the repetitive execution: collecting and sorting results pages, drafting to a brief, and running the checks, across the markets we cover (28 EU markets, the UK and the US).
The cost of a new market falls because the repetitive work doesn’t scale with headcount. The specialists’ time goes where it can’t be replaced: deciding what the market wants. You can see how that’s set up on the NON.agency project page.
If your expansion stalls because you can’t hire fast enough, that’s the resource barrier. Bring us the hard version.
The bottom line
Translating a page and adding hreflang labels a copy. It doesn’t enter a market. The market is entered by rebuilding the piece for its intent, its format, its product language and its sources.
For a challenger, the question is whether you can afford to do that in every language. That’s an infrastructure question before it’s a content one.
What’s still unknown is how AI answer engines weigh local sources against global ones when they build an answer; I’d treat anything more specific than that as a guess. I wrote about how getting named in AI answers works, and you’ll find the wider thread under international SEO. If you’d rather hear me argue it out loud, I host the Polish-language Globalne Horyzonty podcast, and there’s more about me and the work on the About page.
Questions people ask
Is translating a page enough for international SEO?
No. A translation keeps the intent, format and sources of the original market. Search intent and the format that wins the results page differ by language, so the piece has to be rebuilt for each market.
What does hreflang do?
Hreflang tells search engines which language or region version of a page to show to which user. Google says it helps point users to the most appropriate version. It doesn't improve the content itself.
Does Google use hreflang to detect a page's language?
No. Google's documentation says it doesn't use hreflang or the HTML lang attribute to detect language; it uses its own algorithms on the visible content.
How is AI used in international SEO content?
As infrastructure for the repetitive execution, not as a translation shortcut. Specialists decide the format and judge the result, while AI-native systems handle the volume across markets.
