You can, and you should not stop there. Machine translation is now good enough to be a genuine first draft, which is a real change. It is still not good enough to be the final version of the pages that ask people for money. The difference between those two jobs is the whole article.
We get asked this constantly by founders expanding into a second market. The pitch is appealing. Click a button, get a Spanish site, sell to Spain. The reality is more interesting than either the sales pitch or the snobbery about machine translation.
Here is how we think about it, what the platforms actually give you, where Google draws its line, and the workflow we would use on a real project.
Because people buy in the language they think in. CSA Research, in its "Can't Read, Won't Buy" study of 8,709 consumers across 29 countries, found that 76% of online shoppers prefer to buy products with information in their own language, and 40% will never buy from websites in other languages.
That second number is the one that should get your attention. It is not a preference, it is a hard stop. Four in ten people in that study simply will not transact in a language that is not theirs, no matter how good your product is.
The same research found that 75% of respondents are more likely to buy from the same brand again if customer care is in their language, which extends the point past the sale. Language affects retention, not just acquisition.
So the business case is usually settled before the technical conversation starts. The real question is never whether to translate. It is how much to translate, and how carefully.
Good enough that a native speaker will usually understand every sentence, and not good enough that they will trust every sentence. Fluency arrived years ago. Judgement did not. Modern systems produce natural-sounding text and still make confident errors that a monolingual reviewer cannot see.
We deliberately avoid quoting a quality percentage here, because those numbers depend entirely on the language pair, the subject matter, and the type of writing. A machine handles a product specification far better than it handles a tagline built on a pun.
The reliable test is not a benchmark, it is a person. Have one native speaker who knows your industry read the translated homepage and the pricing page. In our experience that single hour surfaces more real problems than any automated quality score.
This is the same position we take on machine-written English, which we set out in our piece on whether you should let AI write your website copy. The tool drafts. A person decides what ships.
Not for using machines. For publishing content with no added value. Google's spam policies name translation directly as one of the automated transformations that can trigger scaled content abuse, describing the problem as generating "many pages (including through automated transformations like synonymizing, translating, or other obfuscation techniques), where little value is provided to users".
Read what that sentence actually targets. The trigger is scale plus emptiness, not the tool. A carefully translated, human-reviewed Spanish version of your service page provides real value to Spanish speakers. Ten thousand auto-generated city pages in twelve languages do not.
Google's broader position on automation is consistent with this. It objects to "using generative AI tools or other similar tools to generate many pages without adding value for users". The value test applies whether the machine wrote the page or translated it.
So the practical rule we follow is simple. Translate pages a real customer in that market would want. Review them. Do not spin up language versions of pages nobody asked for just because the platform makes it cheap.
Webflow Localize includes machine translation as a first pass. According to Webflow, you can translate a field by hovering over it and clicking the globe icon, and the content is translated from your primary locale into the current locale's language. Availability includes page name, SEO title, SEO description, and Open Graph fields.
Two features make it more useful than a raw translate button. You can set a target language's tone and formality, which matters enormously in languages that distinguish formal and informal address. And you can build a custom machine translation glossary for terms you use often, such as product and brand names.
That glossary is the feature teams underuse. Machine translation loves to translate your product name into a common noun. A glossary entry stops that happening on every page, permanently, rather than you catching it by hand each time.
Webflow states that a free preview of basic Localize functionality, including machine-powered translation and static page, CMS, and SEO localization, is included in all Site plans from Starter to Enterprise. That makes experimenting cheap. We wrote about the wider setup in our guide to building a multilingual website in Webflow.
Four things, consistently. Brand and product names get translated into ordinary words. Idioms come through literally and land strangely. Formality levels drift, so one page addresses the reader formally and the next does not. And calls to action lose their punch, because a translated verb is rarely the verb that market actually uses on buttons.
The button problem is the one that costs money. "Get started" has a natural equivalent in most languages, but the natural equivalent is often not the literal translation. A native speaker will tell you the right word in five seconds. A machine will give you a correct word that nobody clicks.
Legal and pricing text deserves separate treatment entirely. Currency formats, date formats, tax wording, and refund terms are not translation problems, they are localisation problems, and getting them wrong creates real liability rather than an awkward sentence.
There is also the failure mode nobody notices: confident nonsense. A translation can read perfectly while saying something you never claimed. This is the same reason we do not publish AI-assisted content without checking it, which we set out in our piece on fact-checking AI-generated content before you publish.
Tier your pages by risk, then apply effort accordingly. Machine translate everything as a first pass, because a full draft is a better starting point than an empty locale. Then treat the review stage as three separate tiers rather than one big job.
The top tier is your homepage, pricing, and any page with a form or a purchase on it. These get a full human review by a native speaker who understands the offer. This is where the money is, and it is usually five to ten pages, not five hundred.
The middle tier is service and product pages. Machine translation plus a glossary plus a skim read by a native speaker. You are looking for brand names, button text, and anything that sounds like it was written by a foreigner.
The bottom tier is your blog archive and older support content. Machine translation is genuinely fine here, reviewed opportunistically. A useful article a customer can read imperfectly beats a perfect article that does not exist in their language.
Translate them deliberately, because they are what search engines and answer engines read first. A translated page with an English title tag and an English slug is a half-finished page, and it will underperform against local competitors who did the whole job.
Keyword research does not translate either. The literal translation of your best English keyword is often not what people in that market actually search for. This is a research task in the target language, not a conversion task, and it is the single most skipped step in multilingual projects.
Search engines also need to be told which version serves which audience, which is a technical annotation job separate from the translation itself. Get that wrong and your language versions compete with each other instead of each reaching its own market.
Start with the conversion path, not the archive. The pages a buyer must pass through to become a customer, translated properly, will outperform a fully translated site where every page is mediocre. Depth beats coverage when you are entering a market.
Be honest about support, too. If someone reads your Spanish page and emails you in Spanish, somebody has to answer. A translated site that funnels people into an English-only support experience creates a worse impression than no Spanish site at all.
We also push clients to pick fewer languages than they first want. Two markets served properly beats six served badly, and the cost of a language is not the translation, it is the ongoing maintenance every time the source content changes.
If you already have traffic or customers from a market, yes, and start with five pages rather than five hundred. If you are hoping translation will create demand that does not exist yet, fix the demand problem first. Translation multiplies an offer that works. It does not create one.
The tooling argument is over. Machine translation inside a platform like Webflow is fast, cheap, and good enough to draft from, and the ability to set tone and maintain a glossary removes most of the repetitive errors. What has not changed is that a human who speaks the language still has to read the pages that matter.
If you want help scoping a multilingual build, or a second opinion on whether it is the right move for your business right now, we are happy to walk through it with you. Let's talk. You can find us at phoenix.studio.
Tell us where you want to go. We'll tell you how we'd get you there.