Should I translate my website into other languages for SEO?
Only once your English pages already earn traffic. Translation multiplies what a page already does — a page that ranks and converts in English can win the same query in a language with thinner competition, but a page earning nothing in English will earn nothing in six more languages. AI translation made the copying almost free, which is exactly why the bar moved: the work that pays is picking the one or two markets you can actually support, translating your best pages properly, and wiring hreflang so engines understand the versions are the same page in different languages rather than duplicates competing with each other.
Translation multiplies traffic — it doesn't create it
Founders usually raise this as a content question — should we have a Spanish version? — when it is really a distribution decision. A translated page inherits the demand of the original and none of its authority. If your English page ranks on page one and converts, a good Spanish version is a genuine second bet on a proven asset, often against weaker competition. If the English page sits on page four, the Spanish one starts from the same zero with fewer links pointing at it, and you have multiplied your maintenance rather than your traffic.
What changed recently is only the cost. Producing another language went from a quote and a two-week turnaround to a few minutes, so the constraint moved from can we to should we. Google's spam policies specifically name text translated by an automated tool and published without human review or curation, which means the cheap version of this is also the version most likely to be treated as scaled content abuse. The expensive parts — knowing what people search in that language, and being able to answer a customer who writes back in it — did not get cheaper at all.
When a second language actually pays off
Translate when all four of these are true. If any one is missing, the pages will exist and do nothing, which is worse than not shipping them, because you now maintain them forever.
- Your English pages already rank — you are duplicating something that works, not hoping a translation rescues a page that never landed. The honest test: could you name the English page you would translate first, and its current position?
- You can support a customer in that language — a signup you cannot reply to is a refund with extra steps. Support, onboarding and sales all have to exist in the language, not just the landing page. This is the condition founders skip most often.
- The demand is real in that language — check search volume for the translated phrasing before you commit, not after. Some markets search in English for technical products, and translating those queries into the local language wins you a keyword nobody types.
- A human reads it before it publishes — a native speaker who knows the product, not just the language. Machine output is a strong first draft and a poor final one, and the errors it makes are the confident kind that a non-speaker cannot see.
Four mistakes that make translated pages worthless
| The mistake | What actually happens |
|---|---|
| Auto-redirecting visitors by browser or IP language | Googlebot crawls largely from US addresses with an English preference, so it gets bounced to the English version and may never see the others. Google's own multilingual guidance advises against automatic redirection for exactly this reason — offer a switcher instead. |
| Publishing translations with no hreflang | Engines have to guess which version to serve, and near-identical pages compete with each other for the same query. The visible symptom is the wrong language appearing in results, followed by a bounce. |
| Translating the keyword literally | The word-for-word rendering of an English keyword is frequently not the phrase people type in that language. You rank for something with no searchers while the phrase that has them goes uncovered. |
| Translating the entire site at once | Every future edit becomes an edit in every language. Pages drift, the stale ones are always the translated ones, and nobody catches it because nobody on the team reads them. |
None of these are translation-quality problems. Each one is a plumbing or scoping problem that survives a perfect translation, which is why teams are so often surprised that a well-written localised site produces nothing. If you are still building the English side, SEO for a brand-new website is the more valuable place to spend the same week.
What running eight locales taught us
First-party note: AgentCeres — the AI Growth Officer at agentceres.com — publishes its marketing site in eight languages, and the parts that took real work were never the sentences. English lives at the root and every other language sits under a path prefix. The language annotations are deliberately not the same strings as the URLs: our Chinese pages live at /zh but are annotated zh-Hans, and Brazilian Portuguese lives at /pt but is annotated pt-BR, because the URL token and the language tag answer different questions. We also chose not to auto-redirect anyone by browser language — a first-time visitor gets English and switches if they want — which is the single decision that keeps every version reachable by a crawler.
- Ship fewer languages, properly. Machine translation buys a first draft in minutes and a maintenance obligation that never ends; we later concentrated on a small number of markets rather than treating all eight as equal bets.
- Treat the URL token and the language tag as separate decisions. Serving /pt while annotating pt-BR is correct, and assuming the URL is the language tag is where hreflang quietly breaks.
- Never redirect by browser language. Offer a switcher, remember the choice, and leave every version crawlable.
- Remember that a passing translation check is not a quality check. Ours proves every locale has every key — it cannot tell you whether a single sentence reads well, and only a native speaker can.
FAQ
- Does translating my website help SEO?
- It can, but it multiplies rather than creates. A translated version of a page that already ranks gives you a second shot at the same demand in a market that is often less competitive. A translated version of a page that ranks for nothing inherits that result exactly, plus ongoing maintenance. The deciding factor is the English page's current performance, not the translation's quality.
- Is machine translation bad for SEO?
- Unreviewed machine translation is the specific thing Google's spam policies call out — text translated by an automated tool without human review or curation before publishing. Machine translation with a native speaker editing before publish is a normal, accepted workflow and is how most multilingual sites are built now. The review step is what separates the two, and it is not optional.
- How many languages should I translate my site into?
- Fewer than the number you can technically produce. Start with one, chosen because you can sell and support in it, and give it real pages rather than a translated homepage with English everywhere behind it. A second language done properly beats six done automatically, and every language you add multiplies the cost of every future content change.
- Do I need hreflang if I only have two languages?
- Yes, and two versions is the easiest case to get right. Each page needs to point at its counterpart and at itself, using a valid language code, and the links have to be reciprocal or search engines ignore them. Without it, two translations of the same page look like competing near-duplicates and the engine picks for you. See hreflang for how the annotation is actually written.
Want this done for you?
AgentCeres is a managed AI marketing team — specialists draft the work, you approve what ships. 14-day free trial, from $19/month.