Here is a pattern we see constantly in international SEO audits. A company enters the French market, pays for professional translation of its best-performing pages, and watches organic traffic flatline for a year. The content is accurate. The grammar is impeccable. Almost none of it ranks.
That is the core problem with SEO translation as most companies practice it. Translation preserves meaning, but search engines do not rank meaning — they rank pages against the exact phrases people type. When your target keywords come from a dictionary instead of from search data, you end up perfectly optimized for queries that do not exist.
Why word-for-word SEO translation fails
A search query is not a sentence. It is a habit: a compressed, market-specific shorthand that people learn from each other, from advertising, from the interfaces they use every day. Two phrases can mean exactly the same thing while one gets searched constantly and the other never gets typed at all.
Four forces drive the gap between translated keywords and real ones:
Commercial vocabulary diverges from dictionary vocabulary. The everyday word for a thing and the word people use when buying that thing are often different. A translator reaches for the everyday word. A buyer types the commercial one.
Loanwords cut across languages unpredictably. French keeps some English words wholesale, rejects others, and invents pseudo-English words that no native English speaker uses. There is no rule a translator can apply; only search data shows which version won.
Register mismatch. Professional translators write polished, formal prose, because that is what they are paid for. Searchers type casual shorthand and abbreviations. The better the translator, the wider this gap tends to get.
Intent hides in different words entirely. Some purchase intents have a standard query word in one language and no equivalent in another. If you only translate your existing pages, you never build pages for demand that exists solely in the target market.
None of this is a translation problem, which is why hiring a better translator does not solve it. It is a research problem.
How French searchers actually phrase commercial queries
Concrete examples make the point better than theory. Every one of these is a head-term pattern in a large commercial category, not an edge case.
Cheap flights. French searchers type "vol pas cher". A faithful translation would give you "vols bon marché" or "vols abordables" — grammatically perfect, commercially dead. "Pas cher" is the workhorse price modifier across French commercial search, where English spreads the same intent across "cheap", "affordable", "budget" and "best value".
Car insurance. The French market says "assurance auto", not "assurance voiture", even though "voiture" is the everyday word for car. Every major French insurer uses "auto" in its product naming, and search behavior follows.
Used and second-hand. For cars, the commercial term is "d'occasion" ("voiture d'occasion"). In fashion, "seconde main" has become the standard phrasing. Same concept, different modifier depending on the category. A dictionary cannot encode that; a keyword tool shows it immediately.
Running shoes. French searchers commonly type "chaussures running", keeping the English word. A conscientious translator will "correct" this to "chaussures de course" and quietly cut the page off from how the market actually searches. Sport and tech are full of these embedded anglicisms.
Email marketing. French marketers search "emailing" — a pseudo-anglicism that is not idiomatic English at all. You cannot arrive at it from either language's dictionary. It only shows up in the data.
Requesting a quote. French buyers type "devis": "devis assurance", "devis déménagement", "devis fenêtre". It is one of the strongest commercial intent signals in French search, and it has no single English twin used the same way. An English site translated into French will typically have no devis pages at all, leaving some of the highest-intent demand in the market to competitors.
France is not the only French. France says "smartphone"; Quebec says "téléphone intelligent". One "French" keyword list cannot serve both markets, which is a preview of a larger truth: keyword research is per market, not per language.
Stack these up and the conclusion is hard to avoid. A translated keyword list will be wrong on the price modifier, wrong on the product term, missing the anglicisms, missing the intent words, and blind to regional splits — all at once. And that is before you account for the fact that French SERPs have their own competitive dynamics and local players; we cover that landscape in our guide to SEO in France.
The right method: research first, transcreate second
The fix is not more careful translation. It is a change in sequence: search data comes before language work, and constrains it.
Run keyword research in the target market before touching any content. Seed it from what actually ranks in the local SERPs: local competitors' page titles, autocomplete suggestions, related searches, category names on leading local e-commerce sites. Your English keyword list can suggest topics to investigate, but it should never supply the phrasing.
Build the French keyword map as its own asset, not a mirror of the English one. Expect gaps in both directions. Some of your English money pages will have little or no French demand behind them. Some French demand — devis queries are the classic case — will have no English source page at all, which means the localized site needs pages the original site never had.
Read the SERPs, not just the volumes. The same query can carry different intent in different markets. Before committing a page to a keyword, look at what Google actually ranks for it in France: commercial pages, informational guides, or something else entirely. The SERP tells you what kind of page you need to build.
Then, and only then, produce the content — as transcreation, not translation. The writer's job is to create a native page that serves the mapped intent and uses the mapped vocabulary, borrowing from the source page where it helps and departing from it where it does not. That includes restructuring sections, swapping examples, and adapting proof points to what a French buyer finds credible.
When our SEO team helped BMW Europe Moto, the official distributor of BMW motorcycle parts, open six new markets organically — work that generated more than 55,000 SEO visitors — the operating assumption held everywhere: six markets means six keyword landscapes, not one list run through translation six times.
Keyword research tells you what to write. The structural layer — hreflang, URL strategy, choosing between ccTLDs and subfolders — decides whether the right page even ranks in the right country. That side of the problem is covered in our multilingual SEO guide.
Briefing translators with search intent
Most localization failures are briefing failures. Translators are trained to optimize for fidelity and elegance; unless you give them a different objective, they will apply it — and "improve" your keywords out of existence.
A useful per-page brief contains:
The primary query, exactly as typed, plus the main variants worth covering. Not a topic, the actual string.
Locked terminology. The terms that must appear verbatim because the data says so, flagged explicitly as non-negotiable. This is what protects "chaussures running" from being corrected into oblivion.
Banned terms. The literal translations you already know are wrong, listed so nobody reintroduces them.
The search intent and page role. What the searcher is trying to do, what the page must help them do, and where it sits in the funnel. This licenses the writer to restructure rather than mirror the source.
Title and meta description requirements, including the target query placement and length constraints.
Internal linking anchors, so localized pages link to each other with keyword-consistent anchor text instead of improvised phrasing.
A brief like this changes the job description. You are no longer asking someone to translate a page; you are asking a native-speaking writer to hit a defined search target using a source page as raw material. That is transcreation, and it is worth paying for. Note that the same logic extends past content into UX, trust signals and conversion — we go deeper on that in our guide to website localization for the French market.
Measuring whether your SEO translation worked
Localized content earns its keep or it does not, and you should decide which in advance. Four measurements matter most:
Non-brand organic traffic, per market. Brand traffic flatters every launch, because people who already know you will find you regardless. Non-brand traffic is the honest signal that your pages are winning queries from strangers. Segment it by country, not just by language.
Rankings on the mapped money terms, checked in the local SERP with the right country and language settings. Rankings checked from your home market are fiction.
Revenue or leads attributed to organic, per market. Traffic that does not convert usually means the intent mapping was wrong, not the writing.
Indexation health. Localized sections often ship with hreflang errors or thin pages that never get indexed. If Google is not indexing the pages, no keyword decision matters yet.
The upside of getting this right is not marginal. When we ran SEO for Gadero, a European e-commerce retailer of outdoor wood products, non-brand organic traffic grew 105% and SEO-attributed revenue grew 59% — and non-brand is precisely the number a translated-keyword approach fails to move, because translated keywords do not intersect with what strangers type.
Be patient on the timeline. Organic results in a new market build over months, not weeks. Set the baseline before launch, review quarterly, and treat underperforming pages as mapping errors to re-research, not as copy to re-polish.
Frequently asked questions
What is SEO translation?
SEO translation is the practice of localizing web content so that it ranks in the target market's search engines, not just reads correctly. Done properly, it starts with keyword research in the target market and uses translation or transcreation to produce pages that match real local queries. Done badly, it means translating existing pages and keywords word for word — which produces content optimized for phrases nobody searches.
What is the difference between SEO translation and transcreation?
Translation reproduces the source text's meaning in another language. Transcreation reproduces the source text's purpose: the writer is free to change structure, examples and phrasing to achieve the same commercial outcome for a different audience. For search content, transcreation constrained by local keyword data is almost always the right model, because the vocabulary that ranks rarely survives literal translation.
Can I just machine-translate my site for SEO?
Machine translation has the same blind spot as human translation — it optimizes for meaning, not for search behavior — and adds quality risks on top. It can be a useful drafting tool inside a process where humans hold the keyword map and review every page. As a publish-and-forget strategy for commercial pages in a competitive market, it forfeits the vocabulary battles described above on every page at once.
Do I still need hreflang if my keywords are localized?
Yes. Keyword localization and hreflang solve different problems: one makes each page match local demand, the other tells Google which language and country version of a page to serve to which users. Google documents the implementation in its Search Central guidelines on localized versions. You need both, and neither substitutes for the other.
Before you translate another page
The order of operations is the whole game: market keyword research first, keyword map second, transcreation brief third, writing last. If you already have a translated site that is not ranking, the diagnosis usually takes one afternoon with the French SERPs — and the fix is a re-mapping exercise, not a rewrite of everything.
Junto's SEO team lives in French search results every day, across 200+ published case studies. If you are weighing an entry into the French market or trying to rescue a translated site, talk to our team — we are happy to look at your keyword map before you spend another euro on translation.
Founder and CEO of Junto
Founder & CEO of Junto, Étienne has been an entrepreneur and digital marketing consultant for over 15 years. An expert in Paid Media, SEO, Data, Automation, AI, Growth and Performance, he helps ambitious companies build high-impact growth strategies — generating lasting results and helping brands move forward in a constantly evolving digital environment.
