Is a multilingual website worth it for a small business?

Adding languages does not add markets — it adds copies. What actually happens in search when you publish the same twelve pages in twenty-two languages, and the three conditions that make a second language pay.

Key takeaways

• Twenty-two languages usually means twenty-two copies of the same pages, not twenty-two markets.
• When versions say the same thing, search engines pick one and treat the rest as duplicates. You publish ten pages and one gets indexed.
• Every language multiplies maintenance: one price change becomes ten edits, and versions drift apart within months.
• A second language pays when you already get enquiries from that country, someone can handle the conversation in it, and the content is rewritten rather than translated.
• Machine-translating your existing pages is the version that reliably fails.

Is a multilingual website worth it for a small business?

Rarely at the start, and almost never in the way it gets sold to you. The pitch is that each language opens a market. What actually happens is that each language opens a copy — of the same offer, the same proof and the same three service pages, now in a language nobody at your company can answer the phone in.

A market opens when someone can reply to an enquiry in that language, issue an invoice into that country and say something specific about it. Translation is the last step of entering a market, not the first.

What actually happens when you publish in ten languages

Three mechanics decide the outcome, and none of them is about how good the translation is.

• Duplicate selection. Search engines compare pages that carry the same structure and the same claims. When several versions are near-identical in substance, one is chosen as the canonical version and the others are set aside. This is not a penalty — it is deduplication working as designed — but the effect is that most of what you published never enters the index.
• Crawl budget. A crawler spends a finite amount of attention on a site of your size. Spread that attention across ten times more URLs and the pages you actually care about get visited less often, not more.
• hreflang has to be honest. The tags must be reciprocal and must point at genuine equivalents. Claiming a language version that does not exist, or pointing at a listing page instead of the article, sends contradictory signals about the same address — and contradictory signals get ignored.

The version of this that hurts most is the quiet one: everything looks fine in your CMS, the sitemap is full, and almost none of it is in the index.

The cost nobody quotes you

The build is quoted once. The maintenance is forever, and it multiplies.

Change your prices and you have ten edits, not one. Publish an article and you have ten publications or one article and nine gaps. Update a legal page after a regulation changes and you have ten chances to miss one. Within six months the versions disagree with each other, and you find out when a customer quotes the old price back at you in a language you cannot read.

This is the part that turns a multilingual site from an asset into a liability: not the initial cost, but the standing obligation.

When a second language genuinely is worth it

Three conditions. If any one is missing, wait.

• Evidence, not hope. You already receive enquiries, orders or traffic from that country. Not "we could sell there" — you already are, awkwardly, in English.
• Someone can carry the conversation. A translated page that produces an enquiry you cannot answer is worse than no page, because you have now disappointed a real buyer.
• The content is rewritten, not translated. Different examples, different regulations, different price expectations, different objections. A German buyer asking about accessibility obligations needs a different answer than a Polish one, and both know it immediately if they are reading the same paragraph in two coats of paint.

Add a fourth if you want the honest version: you are willing to maintain it for years, not launch it once.

Where machine translation works and where it fails

It works well for one job: helping someone who has already found you read what you wrote. If a customer lands on your page and switches it to their language to understand your offer, that is a good use, and nobody is being deceived.

It fails as a growth strategy, and it fails for a specific reason. Search engines and AI assistants reward specificity — a passage that answers a real question with real detail. Machine-translated boilerplate has no specificity to carry across; it arrives as a slightly stiffer version of a page that was already generic. There is nothing in it worth quoting, so nothing quotes it.

What we did on our own site

Worth saying plainly, because it is easy to give this advice while doing the opposite. This site ran in twenty-two languages, with genuinely written articles in five of them. In July 2026 we narrowed the indexed set to those five.

The other seventeen still work. Anyone who lands on them gets a complete, usable page. We simply stopped asking search engines to treat them as separate destinations, because we did not have separate things to say. That is not a retreat from international work — it is matching what we ask for to what we actually have, which is the whole argument of this article applied to ourselves.

The alternative that usually wins

One language done properly, plus three things most multilingual sites forget:

• Say which markets you serve in plain text. "We work with clients across the EU, remotely, in English and Polish" answers the question a foreign buyer actually has, which is not "do you have a page in my language" but "will you work with me".
• Be explicit about currency, VAT and delivery across borders. This removes more friction than translation does.
• Make the contact route obvious and answer it fast. Speed beats localisation at this size.

Then add the second language once the first one is fully worked — when you have run out of things to improve in it, not before.