Skip to content

[blog] Multilingual research · Methods

Researching across languages: why local-language sources change the answer

Searching only in English skews research on non-English markets. Why local sources matter, how to search them well, and the pitfalls to avoid.

The Divrite team 6 min read

Suppose you need to understand the rules for importing electric vehicles into Brazil, the incentives for battery plants in Thailand, or the investment screening process in Germany. You search in English, find several well-written summaries, and assemble an answer.

The answer may well be wrong, or at least out of date and missing the detail that matters. For questions about a specific country, the authoritative sources are usually in that country’s language: the law itself, the regulator’s guidance, the official statistics, the local business press. English-language coverage is generally secondary. It summarizes, it lags, and it covers only what international readers are expected to care about.

This article explains what changes when you search in local languages, how to do it without being fluent, and the mistakes that most often trip people up.

What English-only research misses

Primary documents. Laws, decrees, regulatory notices, court rulings and tender documents are published in the official language. An English translation, if one exists, may be unofficial, partial or several versions behind.

Detail. Summaries tend to keep the headline and drop the conditions: the threshold below which a rule doesn’t apply, the transition period, the exemption for a particular product category. For many practical questions, the conditions are the answer.

Timing. Changes are announced locally first. International coverage, when it comes, may arrive days or weeks later, and some changes never get covered in English at all.

Smaller players. Regional companies, local associations and mid-sized suppliers often have no English presence. If you only search in English, the market looks more concentrated than it really is, because you only see the companies that bother to publish in English.

Local debate. Opposition, implementation problems, and how a policy is actually being enforced tend to be discussed in the local press and in professional forums, not in international summaries.

None of this means English sources are useless. International organizations, global research firms and multinational companies publish valuable comparative data in English. The point is that, for a question about a specific country, English sources should be one input among several, not the whole picture.

How to research in a language you don’t speak

You don’t need to be fluent. You need a method.

1. Decide which languages the question needs

List the countries your question involves, and add any language whose speakers are major stakeholders. A question about a foreign automaker’s plant in Brazil involves Portuguese, but also the automaker’s home-country language, because that’s where its own plans and its home government’s position are discussed in most detail.

Literal translations make poor search queries. Find the term the locals actually use:

  • the official name of the law, program or agency;
  • the regulator’s own name for the concept;
  • the industry’s trade vocabulary.

Good places to find them: the regulator’s website, the official gazette, the relevant article in that language’s Wikipedia (reached through the language links on the English article), or a local industry association’s glossary.

3. Search where the primary sources live

Restrict searches to official domains when you’re looking for primary documents. Many governments use a recognizable domain, such as gov.br in Brazil, go.th in Thailand or bund.de for German federal bodies. Add the official term and a year, and you’ll often land on the source document within a few results.

For business and market questions, look for the local business press, industry associations and stock exchange filings, which are often only published in the local language.

4. Read with machine translation, but verify what matters

Machine translation is good enough to read most documents and decide what’s relevant. It is not good enough to rely on for the passages your conclusions depend on. For those, check:

  • Numbers and dates, against the original text.
  • Negations and conditions: “not”, “unless”, “only if”, “except”.
  • Legal modality: whether something must, may or should be done. These distinctions are easy to lose in translation and they change the meaning entirely.

If a conclusion rests on one sentence, have it checked by someone who reads the language, or compare two different translation tools.

5. Cross-check across languages

Compare what the local sources say with what English summaries say. Discrepancies are informative. Sometimes the English summary is out of date. Sometimes it has simplified a condition away. Occasionally the local press has misreported something that the international coverage got right. Where sources disagree, go back to the primary document.

6. Record sources in their original language

In your source log, keep the original title, the URL and the publication date as they appear, alongside your translation. Quote key passages in the original language with a translation next to them. This lets a reader who knows the language check your reading, and it saves you from re-finding the document later.

A short example

Here is how the steps fit together, using a hypothetical question: “What local-content requirements apply to a battery plant in Thailand?”

  1. Languages: Thai for the rules themselves; English for international comparisons; plus the investor’s home language, if the investor is not Thai.
  2. Terms: find the name of the investment promotion agency and the Thai name of the incentive scheme on the agency’s own website, rather than translating “local content” word by word.
  3. Search: restrict to the agency’s domain and the government domain, with the scheme name and the current year.
  4. Verify: for the clauses that set thresholds and deadlines, compare the machine translation with the original numbers and dates, and check every “must” and “may”.
  5. Cross-check: compare with an English-language summary from a law firm or trade body, and note any condition the summary leaves out.

Common pitfalls

False friends and term mismatches. Words that look alike across languages can mean different things, and the same concept can have different scopes in different legal systems. A “subsidy”, an “incentive” and a “tax benefit” may be distinct legal categories in the local system.

Number formats. Many languages use a decimal comma and a period or space as the thousands separator. Large-number units differ too: 万 and 億 in Japanese, 万 and 亿 in Chinese, lakh and crore in India.

Calendars and dates. Thailand’s official calendar counts years in the Buddhist Era, 543 years ahead of the Gregorian year. Japan uses era names alongside Gregorian years in official documents. Day-month order varies by country.

Regional variants. Brazilian and European Portuguese, or Latin American and European Spanish, differ in vocabulary, including legal and commercial terms. Search using the variant of the country you’re researching.

Stale official pages. Government websites sometimes keep superseded versions online without clear labeling. Check the publication or amendment date, and look for a consolidated version of the law.

One-source dependence. It’s tempting, after the effort of finding one good local source, to rely on it entirely. Local sources have their own perspectives and interests, just like English ones. Aim for at least two independent sources on anything important.

When English is enough

Some questions really are well served in English: topics that are global by nature, most academic literature in the sciences, multinationals that report in English, and international bodies that publish primarily in English. The test is simple. Ask where the primary source for your question is published. If it’s published in English, start there. If not, that’s the language your research needs.

How Divrite handles this

Doing this by hand for every question takes time, which is why we built it into Divrite’s deep research. When your question involves specific countries, Divrite searches in their languages as well as in English, compares what the sources say across languages, and cites each source so you can open the original. The report is written in the language you asked in, whichever languages the sources were in.

The checks in this article still apply to anything you rely on, particularly the passages where numbers, conditions and legal obligations live. But starting from the right sources in the right languages makes the rest of the work considerably easier.

Filed under Multilingual researchMethodsAI research
All articles →

Research you can check, in one question

Divrite breaks your question into tasks, searches in many rounds and languages, and ties every figure to a source you can open.