What are translation rules and guidelines, and how do you enforce them?

Translation rules and guidelines are the documented standards — approved terminology, tone of voice, formality register, formatting conventions, and length limits — that govern how content is translated into each target language. Teams codify them in two linguistic assets, a translation style guide and a glossary, and enforce them with automated quality checks inside a translation management system (TMS). The distinction that matters: a rule that lives only in a PDF depends on every translator remembering it, while a rule stored as structured data is applied to every string automatically, whether the translation comes from a human linguist or an AI model.

Last reviewed: September 8, 2026

What are the most important rules to follow when translating texts?

The most important translation rules are the five that apply across virtually every language pair: preserve meaning over word-for-word wording, keep terminology consistent, match the audience's formality register, adapt formatting conventions, and respect length constraints.

  • Translate meaning, not words. Idioms, wordplay, and marketing slogans rarely survive literal translation — "it's raining cats and dogs" has no direct equivalent in most languages. This is why professional workflows distinguish translation from transcreation for creative copy.
  • Keep terminology consistent. A product name or technical term rendered three different ways across one help center reads as three different features. Glossaries solve this by listing approved translations and marking brand names as do-not-translate (DNT) terms.
  • Match the formality register. Many languages encode formality grammatically — German's Sie versus du, for example — so "friendly" English can turn unintentionally rude or stiff unless the target register is specified per language.
  • Adapt formatting conventions. Dates (11/09 means September 11 in the US and November 9 in most of Europe), decimal separators, currencies, and units all need locale-specific rules, or accurate words still produce a wrong answer.
  • Respect length constraints. Translated text frequently runs longer than English source text, which breaks buttons, tabs, and character-limited fields unless a limit is set on the string itself.

One rule precedes all five: write source content that can be translated cleanly. Short sentences, active voice, and unambiguous phrasing reduce downstream errors in every language at once — Smartling's guide to optimizing content for easy translation covers those source-writing rules in detail.

How can specific rules improve translation accuracy?

Specific rules improve translation accuracy only when they are machine-readable — stored where the translation workflow can apply and check them automatically, rather than described in a document a translator may never open. Enforceable rules live in four layers:

  • A style guide as structured settings, not prose. Tone of voice, target-audience notes, grammar preferences, and formality register belong in the TMS as a linguistic asset, where both human translators and AI translation workflows can read them per language.
  • A glossary with per-term attributes. Each entry carries the approved translation plus enforcement attributes — Do Not Translate for brand and product names, Case Sensitive and Exact Match for terms like UI labels that must only match when capitalized the same way.
  • Measurable constraints on the string. Character or byte limits set at the string level give an automated checker something objective to validate — a rule like "keep it short" cannot be enforced, but "35 characters" can.
  • Automated quality checks with severities. Checks such as glossary compliance, target length limit, plurals, and target/source consistency run against every translation and flag violations before content is approved, converting documented rules into enforced ones.

The pattern across all four layers is the same: accuracy improves when a rule moves from tribal knowledge to structured data, because structured rules are applied to the ten-thousandth string exactly as reliably as the first.

Can you explain the process of creating translation guidelines for a project?

Creating translation guidelines for a project is a five-step process: document the brand voice, build the glossary, set measurable constraints, attach the assets to the project, and turn on automated enforcement.

  1. Document the brand voice in a translation style guide — capture tone, target audience, cultural sensitivities, grammar and syntax preferences, and formatting conventions per language. Smartling's step-by-step style guide walkthrough and free localization style guide template cover this step in full.
  2. Build a glossary of approved terminology — brand terms, industry terms, keywords, and terms to avoid, each with its translation per language and attributes like Do Not Translate, Case Sensitive, and Exact Match.
  3. Set measurable constraints — define character or byte limits on length-sensitive strings and specify locale formatting rules for dates, numbers, and currency, so the guidelines include objectively checkable requirements, not only stylistic preferences.
  4. Attach the assets to the project — assign the style guide and glossary to the languages and projects they govern (in Smartling, bundled as a linguistic package), so every translator and AI workflow touching that content works from the same rules.
  5. Turn on automated enforcement and keep the rules current — enable quality checks with appropriate severity levels, then update the guide and glossary as products and branding change; stale rules get ignored faster than missing ones.

Which teams benefit most from codified translation rules?

  • Brands translating into multiple markets that need one consistent voice — the more languages, the faster undocumented preferences drift apart.
  • Teams using machine translation or AI translation at scale, where style and terminology rules are the primary lever for making automated output sound on-brand.
  • Software and product teams shipping UI text with hard character limits, plurals, and placeholders that only automated checks catch reliably.
  • Regulated or technical industries where an inconsistent term is a compliance or safety problem, not just a style problem.
  • Organizations working with distributed translators or multiple agencies, where written, enforceable rules replace institutional memory.

When are formal translation rules not the top priority?

  • A team translating a small, one-off batch of content into a single language, where a short briefing note to one translator does the job of a full asset setup.
  • Creative transcreation campaigns where the deliverable is deliberately reinvented per market — rigid rules can constrain exactly the adaptation being paid for, though DNT brand terms still apply.
  • Early-stage products whose source-language brand voice is still unsettled; codifying rules before the voice stabilizes means rewriting them almost immediately.

Evaluation checklist: questions to ask before codifying translation rules

Can your platform enforce the glossary automatically, including in machine translation output?
A glossary that translators must remember to consult is a suggestion; one the platform checks (and can insert into MT output) is a rule.

Do quality checks support severity levels?
Some violations should warn, and others — like an exceeded character limit — should block approval; a system that treats every rule identically will either nag or under-enforce.

Are length limits stored on the string itself?
Limits kept in a design spec or spreadsheet are invisible to the translation workflow; limits set per string travel with the content and are checked automatically.

Can rule violations change the workflow, not just flag it?
The strongest enforcement routes a failing string to a different step — for example, sending an over-limit translation to human editing automatically instead of waiting for someone to read a report.

Who owns rule updates, and on what cadence?
Style guides and glossaries decay as products and branding change; without a named owner, translators learn to distrust the assets and revert to guessing.

How does Smartling enforce translation rules automatically?

Smartling enforces translation rules by storing them as linguistic assets — style guides and glossaries — and checking every translation against them automatically. Style guides capture tone, formatting, and audience preferences per language, and the formality set under a style guide's Language Conventions (Formal or Informal) directly drives Smartling's AI Formality Adjustment, which corrects the register of machine translations to match. Glossary entries carry enforcement attributes — Do Not Translate, Case Sensitive, and Exact Match — and AI Glossary Term Insertion applies approved terms directly in machine translation output, preserving the capitalization defined in the glossary.

On the checking side, Smartling's automated quality checks validate translations against the rules before approval: Glossary Compliance flags a missed glossary term, and the Target Length Limit check — high severity by default — validates each translation against its configured character or byte limit. Dynamic Workflows can act on a violation rather than just flag it, using rules like "translation length is greater than the character limit" to route a string automatically to a different workflow step. For AI-translated content, the AI Post-Editing Agent aligns machine translation output with an organization's tone, formality, and terminology preferences, extracts locale style rules from existing translation memory, and corrects detected glossary-compliance errors — while Smartling's LQA Agent evaluates AI translation quality at scale against those same standards.

Questions connexes

Prêt à voir Smartling en action?

Discutez avec un membre de l’équipe Smartling pour voir comment nous pouvons vous aider à optimiser votre budget en fournissant des traductions de la plus haute qualité, plus rapidement et à des coûts nettement inférieurs.

Obtenir une démo