Why translating your site into French never makes it rank

Every few months I look at a site that has done everything right in English and cannot understand why France is flat. The French pages exist. They are indexed. Core Web Vitals pass. Search Console shows impressions and a click line that never moves. The conclusion in the room is always the same: France is a difficult market.

France is not a difficult market. The keyword list was wrong before anyone wrote a single page, and nothing downstream can fix that.

The mechanism, in one sentence

A translator's job is to be accurate. A searcher's behaviour is to be lazy. Translation gives you the most correct French term; ranking requires the most typed one. When those diverge, and they diverge constantly, translation quietly aims your entire French section at the wrong words.

The numbers, so you do not have to take my word for it

Three pairs, all measured on 30 July 2026 through Google Ads keyword data (country France, language French). In each row, both terms are defensible translations of the English concept. Only one is a market.

Look at the second row for a moment. The full French phrase for search engine optimisation, *optimisation pour les moteurs de recherche*, is what a careful translator produces. It gets 590 searches a month. French people overwhelmingly type the English acronym: SEO, 27,100. Being more French than the French market cost you 98% of the demand.

The third row is the subtle one. *Agence de référencement* is real, established, and gets a healthy 2,900 searches. It is not a mistake. It is just the smaller of two live markets, and if you only build for it you left the bigger one to a competitor.

Why nobody catches this for two years

Because every diagnostic you would normally run comes back clean. That is the genuinely dangerous part.

• Indexation: fine. The pages are in the index.
• Technical SEO: fine. Same template as your English pages, which pass.
• Rankings: fine, and this is the trap. You do rank, often in the top five, on terms with almost no volume.
• Traffic: flat. Which reads as a demand problem rather than a targeting problem.

Diagnose it yourself in ten minutes

You do not need a consultant to find out whether you have this problem. You need Keyword Planner and ten minutes.

1. List the five terms your French pages actually target. Take them from your French H1s, not from your strategy deck.
2. Open Google Ads Keyword Planner. Set the location to France and the language to French. Both matter: leaving the language on English returns a different and useless picture.
3. Enter your five terms and note the volumes.
4. Now enter what you suspect a French person would type instead, including any English loanword. French uses more English tech vocabulary than most languages, and *SEO*, *marketing* and *design* are all normal French usage.
5. Compare. If any pair differs by more than about three times, your French section is built on the wrong term.

Write today's date next to the numbers. Search volume moves, and a figure without a date stops being useful within a year.

The fix is an order of operations, not a tool

There is no plugin for this. The correction is to move one step earlier in the process and never skip it again.

1. Research in French, from your categories. Start from what you sell, not from your English keyword list. Starting from the English list guarantees you inherit its blind spots.
2. Pick on volume plus intent, not on elegance. If the ugly loanword is what people type, it goes in the H1. Your brand voice can survive it.
3. Brief in French. The writer receives the target term, the volume and the intent, not an English page to convert.
4. Write in French. Not translate into it. French buyers expect different proof and different reassurance, and a translated page reads as translated in about two seconds.
5. Then wire hreflang. Both directions, plus a correct x-default. Reciprocity is what breaks most often: if your French page points at the English one and the English one does not point back, Google discards the pair and picks a version itself.

The same mistake, now in AI answers

This is about to matter more, not less. When a model answers a question in French, it picks from French sources. A translated page is a weak candidate: the phrasing is slightly off, so it matches the French question less well, and the page usually states no quotable facts because faithful translation strips out the specifics that would have been added natively.

The upside is that French AI answers have far fewer credible sources to choose from than English ones, so the bar is lower while it lasts. That is a separate piece of work, covered on AI search optimisation.