Multilingual SEO: Why Keywords Are Not Translated Directly

You can translate a page flawlessly and still find that nobody in the target market can locate it. That is because the phrases users type into a search box are not translations of the phrases in the source text. A different language means a different search behaviour.

Contents

Multilingual SEO is part of search engine marketing, and its aim is to increase visitor traffic, reach more customers or raise an organisation’s profile (Jiménez-Crespo 2024, p. 186). Within localisation it is the area that sits closest to the marketing and business side.

Below we set out why this work is not translation alone, which tags carry visibility, how the preparation is done and where the technical side comes in. The aim is not to second-guess the algorithm; it is to make clear which decisions are taken at the translator’s desk and which at the developer’s.

Where does multilingual SEO sit within localisation?

In the localisation process, SEO is one of the last items in the preparation stage. The source counts the planning of web layout, indexing and search engine optimisation as a separate step within preparation; keywords and descriptions are prepared here so that the page is indexed correctly in the target market (Jiménez-Crespo 2024, p. 185).

The work calls for a different set of skills. Chief among them are knowing how users behave linguistically, that is, how they search; translating keywords and carrying out comparative analysis across languages; writing translations under SEO constraints; and producing off-site copy in the target language (Achkasov 2015).

Search engines keep their ranking algorithms secret. What is known, by contrast, is which elements are read and how they should be formatted. There is no exact formula, then, but there are workable rules. Setting out that distinction from the start is what makes a realistic expectation possible.

Why is a keyword not translated directly?

Keywords and titles translated into the target language word for word are largely ineffective (Jud and Massey 2011: 148). The reason is simple: the accuracy of a translation and its search volume are different things. A phrase that is correct but that nobody types produces no visibility.

The recommendation, therefore, is that the localiser draw up an independent keyword list for the target site before translation begins. The list comes not from the source text but from search behaviour in the target market. It is a separate research step that sits ahead of the translation work.

Once the list is ready, the relevant components have to be flagged in the translation memory tool and dedicated adaptation strategies applied to them. More often than not, working alongside local linguists and local marketing teams becomes unavoidable.

Which tags carry visibility?

Search engines do not read every part of a page with equal weight. According to the source, the most important fields are the metadata tags, the page title, the subheadings, the keyword field and image alternative texts (Jiménez-Crespo 2024, p. 187). The table below shows these and what they correspond to in localisation.

Element	Where it appears	Localisation approach
Page title	Browser tab and search result heading	Adapted; kept short so that it is not truncated
Description metadata	Beneath the heading in the search result	Rewritten, not translated word for word
Keyword field	Not visible to the user	Reproduced to suit the target region
Subheadings	Within the page	Key phrase moved to the front
Image alternative text	When the image does not load and in a screen reader	Translated descriptively
Link text	In links within and beyond the page	Written in clear, self-explanatory wording

What the items in the table have in common is that most of them are not directly visible to the user. That is why they are so easily forgotten when the scope of a translation is being set. The text read on screen gets translated; the page’s metadata stays in the source language. The result is a page that is written correctly for the target market but tagged wrongly.

How are the title and the description written?

The page title is the text that appears at the top of the browser window and in search results. The source recommends keeping the title in the 50-60 range so that it is not truncated in the browser or in the search result (Jiménez-Crespo 2024, p. 180). Because the same information usually takes up more room in Turkish, that limit comes under pressure.

The description metadata, meanwhile, is a short introduction to the page and appears beneath the heading in the search result. It tells both the search engine and the user what the page is about. This text does not have to be a translation of the description in the source language; the wording chosen is the one that earns clicks in the target market.

The keyword field no longer carries the weight it once did. The reason is the habit developers had of stuffing the field with unnecessary words in order to mislead the algorithm. If it is used at all, it should be filled with local equivalents rather than with a direct translation.

The technical side: factors beyond the text

SEO does not have a linguistic side alone. The technical dimension concerns the site’s architecture, its coding and the absence of dead links. Site security and user experience are counted on this side too (Jud and Massey 2011). These are the responsibility of the development team, not the translator.

Knowing that distinction matters for managing expectations. Even if the text side is flawless, visibility will not rise while the technical side is faltering. The reverse holds as well. In projects where the two sides work without knowing what the other is doing, the outcome usually disappoints both teams.

Online tools can extract keywords from existing pages. Among these the source counts search trend and advertising keyword tools, along with audit extensions built into the browser. Such tools shorten the preparation stage but are no substitute for the judgement of a local linguist.

Techniques that work in practice

The source lists a set of widely accepted techniques for raising visibility in the target language. These are not magic formulas but plain rules that make the text legible for the user and the indexer alike.

1. Use key phrases that state the subject of the page plainly, preferably made up of more than one word.
2. Place the key phrases in the opening words of every paragraph, section and heading.
3. Keep titles short and do not go past the limit of sixty.
4. Keep links in working order, and use clear, self-explanatory link text for links within the page.
5. Ask the developers for an analysis of the searches that bring traffic to the page in the target market.
6. Observe word limits where they apply; in news copy that limit is often three hundred words.
7. Add the spelling mistakes users commonly make to the keyword list.

Most of the items on this list are rules of good writing as well. Putting the key phrase first fits the fact that readers scan the screen rather than read it. A short title prevents truncation on a narrow display. In other words, most of what is done for visibility improves readability too.

Overdoing it, on the other hand, backfires. Repeating the same phrase in every sentence makes the text feel artificial and drives the reader away. The source puts the keyword field’s loss of value down to exactly this kind of forcing. The measure to aim for is a density that keeps the text flowing naturally.

Images and videos are ranked too

Search engine visibility is not a matter of pages alone. There is ranking in image search and on video platforms as well. If a localised page contains images or video, their titles, descriptions and alternative texts should be adapted before upload (Jiménez-Crespo 2024, p. 188).

Alternative text has a dual function. On one hand it supplies context for visually impaired users and for visitors who turn images off on a slow connection; on the other it helps search engine crawlers make sense of the page. It should not, therefore, be fobbed off with filler text.

Writing alternative text requires knowing what the image is for. For purely decorative images the field is left empty; for charts and tables that form part of the content, the whole of the information is described. Ending the text with a full stop makes the screen reader pause.

It is normal for the same image to carry different alternative texts in different language versions. The file stays the same; the description is localised. The source demonstrates this through a three-language example: the same photograph of a bridge is tagged in each language with that language’s own description. In most projects this detail is skipped as a matter of course.

The Türkiye context

Search behaviour in Turkish creates a problem of its own, because of the way the language is built. Its agglutinative structure means the same concept can be written dozens of ways: “çeviri fiyatı”, “çeviri fiyatları”, “çeviri ücreti”, “tercüme fiyat”. Deriving these variants from a single source phrase is not possible; the list has to be drawn up locally.

The second issue is Turkish characters. A proportion of users type into the search box without circumflexes and without dots. Adding these variants to the keyword list as well chimes with the source’s advice on spelling mistakes. In the address structure, converting Turkish characters is standard practice.

The third is the choice of term. With synonymous pairs such as “Çeviri” and “tercüme”, which form is searched for more often varies by sector and by region. That preference is a market decision rather than a translation decision, and it should be backed by data.

What does this mean for you?

Leaving visibility until later when you open your site up in a new language usually means the text has to be reworked a second time. The order below prevents that.

1. Have an independent key phrase list prepared for the target language before translation starts.
2. Match the list to the pages one by one, assigning a main phrase to each page.
3. Deliver the title and description metadata in a separate file, together with their character limits.
4. Bring image alternative texts into the scope of the translation.
5. Follow the search data over the first three months after going live and update the list.

These steps do not add a great deal to a translation budget, but they make the return on the work visible. You can look over the scope of what we offer on our translation services page, and reach articles on similar subjects from our knowledge centre.

Frequently asked questions

When is the keyword list prepared?

Before translation begins. The source states that the localiser should draw up an independent keyword list for the target language before starting on the text. If the list is prepared afterwards, the text has to be gone over again and the work is done twice.

Does the keyword tag still matter?

Not as much as it used to. The tag has lost weight because developers stuffed the field with unnecessary words in order to mislead the algorithm. If it is to be filled at all, it should be filled with phrases adapted to the target region rather than with a direct translation.

Do images count for anything in a search engine?

They do. Search engines rank images and videos as well. Image alternative texts, video titles and video descriptions therefore need to be adapted into the target language before upload. Alternative text is also necessary for screen readers.

Should a translation agency handle multilingual SEO?

Most projects call for a division of labour. Keyword research and local market knowledge mean working with a local linguist and a marketing team. The translation side works that list into the text; the technical side is the developer’s responsibility.

Sources

• Jiménez-Crespo, M. A. (2024). Localization in Translation. Routledge, pp. 185-188.
• Jud, P. & Massey, G. (2011). Search engine optimisation and translation, p. 148.
• Achkasov, A. (2015). On the skills required in SEO-oriented translation.