A ceramics workshop in Seville, a boutique hotel in Valencia, an online shop selling olive oil from Jaén. More and more small Spanish businesses sell to customers who do not speak Spanish. Their websites need English, their product sheets need German, their booking confirmations need French.
With limited budgets, many owners turn to free machine translation. It is instant, costs nothing and often looks good enough. Sometimes it is. Sometimes it quietly damages the business.
What the engines do well
Modern neural engines have improved dramatically. For common language pairs such as Spanish to English, they produce fluent, grammatically correct sentences most of the time. They are useful for:
- Understanding incoming emails, reviews and supplier messages
- Drafting internal notes and quick replies
- Getting the gist of documents before deciding what to translate properly
- Translating large volumes of simple, repetitive product data
Used this way, a good machine translation tool saves time and money. The key is knowing where to stop.
Where things go wrong
Machine translation struggles with context, tone and anything that depends on culture. A playful Spanish product description can become flat or even odd in English. Idioms get translated literally. Names of dishes, regional products and local expressions are often mangled.
The engine also has no idea who your customer is. It does not know whether to address them formally or casually, whether the text is a legal notice or a marketing slogan, or whether a word is your brand name and should not be translated at all.
Worst of all, it never flags uncertainty. A wrong sentence looks exactly as confident as a correct one.
Languages with a different alphabet
Quality varies by language pair. Results between Spanish and English or French are usually decent. For languages with different scripts and grammar, errors increase. Businesses on the Costa del Sol that attract many Russian-speaking visitors know this well.
A hotel website or a real estate brochure aimed at this audience deserves a proper english to russian translation by a native linguist. Russian has complex grammar and strict conventions for politeness, and machine output often sounds stiff or unintentionally rude to native readers.
Legal and official texts
Some texts should never rely on machine output alone. Terms and conditions, privacy policies, return policies, contracts and anything submitted to an authority carry legal weight. A mistranslated clause can create obligations you never intended.
Official documents are a separate category again. A Spanish company registering a branch in Munich, or an employee relocating to Germany, will need certified german translation services for company deeds, diplomas or civil records. No engine can certify a translation, and German authorities will reject anything that is not signed and stamped by an authorised translator.
A hybrid approach
For many small businesses, the most practical solution is a mix. Machine translation produces a first draft for low-risk content. A professional linguist then reviews and corrects it. This process, known as post-editing, can cut costs while keeping quality acceptable.
High-visibility texts, such as the homepage, main product pages, ads and legal pages, are better translated from scratch by a human. They shape the first impression, and they are usually short enough that the cost stays manageable.
Protecting your data
Free online engines may store the text you paste into them. For customer data, contracts or unreleased product information, that is a real risk. Business-grade tools with clear data policies, or professional translators bound by confidentiality agreements, are the safer choice.
Search engines care too
If your website relies on search traffic, translation quality affects visibility. Search engines can detect low-quality, automatically generated content, and customers searching in English or German use specific terms that a literal translation may miss. Good localisation includes keyword research in the target language.
Measuring the cost of errors
It helps to think about translation errors in terms of lost sales rather than saved fees. A tourist who finds a confusing cancellation policy may book elsewhere. A wholesale buyer who reads a clumsy product sheet may doubt the quality of the goods themselves. A single bad review mentioning a misleading description can undo months of marketing work.
Seen that way, professional translation of a few key pages is not an expense. It is insurance for the reputation you have built at home, now exposed to customers who have never heard of you.
A simple decision guide
- Internal and temporary content: machine translation is usually fine.
- Large volumes of simple product data: machine translation plus human review.
- Marketing, brand and homepage content: human translation.
- Legal, medical and official documents: specialised or certified human translation.
The bigger picture
Machine translation has a long history, from rule-based experiments in the 1950s to today's neural networks, as the Wikipedia article on machine translation describes. It is a remarkable technology, and small businesses should use it.
Just not blindly. The customers you want to win abroad will judge you on every word they read. Spending wisely on the texts that matter most is one of the smartest investments a small exporter can make.