9 types of translation and how to choose the right approach

5 types of translation based on content format 

Not all content translates the same way. 

Error messages in an app need to be crisp and easy to understand. Marketing campaigns should have cultural nuance and resonance. Legal contracts require zero-tolerance accuracy. 

Keep these differences in mind, let’s look at five types of translation based on content formats.

1. Software and product localization

Best for: SaaS platforms, mobile apps, web applications 

Cost-Efficiency: Medium to high efficiency. AI translation can handle most UI strings; human review focuses on critical user-facing elements and context-dependent translations.

Software and product localization adapts digital products, such as mobile apps and web applications, for specific markets and languages. Beyond UI strings, this type of translation covers error messages, in-app notifications, onboarding flows, tooltips, and more. 

Modern teams also rely on design-stage localization to speed up the process.

This approach lets multiple teams (product, design, development, and translation) validate translations against actual UI layouts before development. It prevents costly post-launch fixes.

When should you use software and product localization?

Localize your software when market signals indicate readiness. Some indicators are a growing user base and support tickets in non-native languages. You can also check if competitors in target markets gain traction with localization.

2. Technical documentation translation

Best for: Developer tools companies, B2B SaaS with complex products, enterprise software 

Cost-efficiency: High efficiency. Structured content with consistent terminology can use translation memory heavily.

Technical documentation translation covers documents like:

• Help centers
• User manuals
• Release notes
• Knowledge bases
• Installation guides
• API documentation
• Troubleshooting articles

Technical localization prioritizes clarity, precision, and findability, as users seek this content to solve specific problems. Besides, it involves many industry-specific terms. That’s why this type of translation requires high technical accuracy and consistency.

Share your company’s glossaries and reference materials with technical translators to get solid output. You can also connect them to your in-house experts for complex concepts.

When should you use technical documentation translation?

Use technical documentation translation when your product requires user education and support in multiple markets. 

The best time to invest in this type of translation is when you notice a spike in support tickets in other languages. It’s evident that users struggle with English-only documentation. Another scenario is if regulatory requirements make it essential to share instructions about your product in the local language.

3. Marketing translation

Best for: E-commerce brands, B2C companies, consumer tech products

Cost-efficiency: Low to medium efficiency. Requires transcreation for creative content (expensive), but product descriptions and email templates can use AI translation with light human review.

Marketing translation localizes promotional content, including landing pages, email campaigns, social media posts, and messaging. It tailors your brand experience for every target market while preserving your voice and the original campaign's persuasive impact.

This type of translation sits at the intersection of language, culture, and marketing strategy. Rather than word-for-word translations, it often leans on transcreation, the process of recreating the emotional resonance, cultural references, and impact of the source content.

In other words, marketing translation aims to reimagine your brand while maintaining brand intent and connecting with local audiences. 

When should you use marketing translation?

Marketing translation is ideal when you want to enter markets where cultural connection drives purchasing decisions.

This approach becomes critical for:

• Launching paid advertising campaigns 
• Creating local-language landing pages for conversion optimization
• Building brand presence in markets where English-first messaging falls flat

4. Multimedia content translation

Best for: Streaming platforms, e-learning companies, marketing teams for global brands

Cost-efficiency: Medium efficiency. Subtitling is cost-effective, but dubbing is expensive.

Multimedia content translation focuses on subtitling and voiceover dubbing to localize video content, podcasts, e-learning materials, webinars, and more. 

Unlike other translation types, Multimedia translation must balance linguistic accuracy with strict time and space limitations. Subtitles can only display so many characters per second while remaining readable.

And you have to convey the same meaning in a new language within the exact timeframe of the original audio.

That means you have to consider factors like subtitle timing, reading speed, lip-sync requirements, and accessibility compliance.

When should you use multimedia content translation?

Multimedia translation is worth the investment when video or audio content is a primary channel for customer engagement, education, or conversion for your business. 

Look at your YouTube analytics data to see significant viewership from non-English-speaking countries. Or you can also analyze the leads generated from video campaigns to check if a good chunk of them comes from non-domestic markets. 

5. Legal translation

Best for: Brands in regulated industries (fintech, healthcare, SaaS with enterprise customers)

Cost-efficiency: Low efficiency. Requires specialized translators with legal credentials and multiple expert review cycles.

Legal translation converts binding documents like contracts, terms of service, patent applications, and compliance materials into legally valid translations. It preserves the exact meaning and legal effect of the source content.

Legal translation is one of the highest-stakes translation categories. Any errors can nullify contracts or create liabilities for your business. 

When working on this type of translation, you need translators with extensive legal training in both source and target languages. And your translators must also understand document formatting conventions, as legal documents follow strict structural requirements in different countries.

When should you use legal and compliance translation?

Use legal translation whenever documents have legal force, regulatory implications, or create contractual obligations in markets operating in different languages. 

This is non-negotiable for: 

• Filing patent applications in foreign patent offices 
• Submitting regulatory compliance documentation
• Creating enforceable contracts with international partners
• Establishing legally compliant terms of service and privacy policies 

4 types of translation based on technology

AI-powered translation has become table stakes for scaling your localization efforts. 

But every team and every project needs a different approach to AI translation. We break down the four approaches to AI translation to help you choose the right one for your needs. 

1. LLM-based translation

Best for: Content that needs cultural adaptation and tone matching.

Cost-Efficiency: Medium efficiency. More expensive per word than traditional NMT but delivers significantly better quality for complex content.

LLM-based translation uses large language models like GPT-4o, Claude Sonnet/Opus, and Gemini to generate translations. These LLMs understand the language context, cultural nuance, and intent. 

Unlike earlier machine translation approaches, LLMs don’t rely solely on pattern matching from training data. They can:

• Reason about translation choices
• Adapt tone and formality based on context
• Handle ambiguous phrases by considering the surrounding content

Besides, LLM translation produces a higher quality output by processing entire paragraphs or documents as a unit instead of sentences.

When should you use LLM-based translation?

LLM-based translation is the right choice when your source content’s tone, context, and cultural nuance matter more than speed or cost efficiency. Think marketing campaigns where maintaining voice and emotional impact is critical. 

You can also use LLM-based translation for:

• Technical documentation requiring consistent terminology across related documents
• Customer-facing content where awkward phrasing damages user experience
• Creative materials like social media posts, where literal translation falls flat

LLM vs NMT-based translation

While both LLM-based translation and Neural Machine Translation (NMT) use AI, they’re fundamentally different in their approaches.

Neural Machine Translation (NMT) systems like DeepL and Google Translate learn statistical patterns from billions of translated sentence pairs. For a sentence like “the bank is closed,” NMT tools search their training data and apply the most statistically probable translation. 

However, this doesn’t work as well when you’re working with context-dependent content. For a sentence like “I sat by the riverbank,” an NMT tool might give you the incorrect translation for the word bank. 

On the other hand, LLM-based translation looks at the source content as a whole to create contextually meaningful translations. LLMs apply their reasoning skills to pick the more relevant words and adapt tone according to the context.

The key difference: NMT optimizes for statistical probability across short segments, while LLMs optimize for contextual accuracy across entire documents.

2. AI orchestration translation

Best for: Teams managing diverse content types simultaneously and lean orgs handling high translation volumes

Cost-efficiency: Very high efficiency. Publish-ready translations at a faster turnaround time and high cost savings compared to traditional agency workflows.

AI orchestration is a context-aware translation system that dynamically routes every translation task to the best-suited AI model. 

For example, Lokalise’s AI orchestration engine evaluates each project based on content type, language pair, and available context. Based on this analysis, it selects the right AI engine from multiple models like Claude, GPT, DeepL, Google Translate, and more.

More importantly, Lokalise uses context from translation memory, glossaries, screenshots, style guides, and character limits before choosing the right model. 

As a result, you can maintain high cost efficiency and speed instead of taking a one-size-fits-all approach. 

Lokalise also uses translation quality scoring to assess the accuracy of every output. Translations with a high score of 80 or above are automatically approved. And translations scoring less than 80 are routed to your linguists for further review with suggestions for improvement.

When should you use AI orchestration translation?

AI orchestration is an effective way to scale translation volume without proportionally scaling costs or compromising quality. This translation approach can help you expand to multiple markets simultaneously without expanding the localization team size proportionally. 

Here are other ways when AI orchestration comes in handy:

• Ship product updates at a daily or weekly cadence
• Managing diverse content types that need different quality thresholds

3. Agentic translation

Best for: Lean teams without dedicated localization managers

Cost-efficiency: Very high efficiency. Eliminates project management overhead and prevents costly post-deployment fixes through design-stage validation.

Agentic translation involves AI agents autonomously managing translation workflows through a multi-stage reasoning process. 

Traditional translation tools require manual setup and multi-step configuration. Agentic systems simplify this into an intelligent, automated workflow:

• Understand goal: The agent interprets natural language requests like “localize the new checkout flow” and understands what you're trying to achieve.
• Gather relevant context: The system collects context from translation memory, checks glossaries and style guides, and analyzes UI screenshots to understand where text appears.
• Reason and plan: The agent uses its reasoning skills to make an action plan: which translation engines to use, what quality checks to apply, and how to handle technical constraints.
• Execute the plan: The agent interacts with your translation management system to implement the translation workflow. It can also run quality checks and learn from human feedback to improve future translations.

Agentic translation has a built-in planning and reasoning approach. That means your translation workflows can adapt intelligently to your specific context rather than following rigid, pre-programmed rules.

When should you use agentic translation?

Agentic translation is ideal for lean teams working on a tight budget and without a dedicated localization manager. Agentic workflows can learn your preferences over time. So, you can proactively train the agent with relevant context to improve the quality of translations.

Besides, this type of translation also prevents translation from becoming a bottleneck. 

If your product team releases new features every week, designers can instantly translate the new UI to test their mockups and roll out these features quickly. 

How does an AI agent use translation memory?

AI agents actively use reasoning skills when applying Translation Memory. Instead of directly applying all the 100% matches, agents find segments of text and analyze the context to decide whether these matches are relevant. 

For example, if the term “settings” appears in a different context (like account settings vs. payment settings), an AI translation agent recognizes the difference and adapts accordingly.

Agentic translation workflows can also work with fuzzy matches. Agents identify the difference between the source segment and TM suggestions to automatically modify the translated output.

Through this process, translation agents can learn your translation preferences over time and apply them intelligently across all content

4. Custom AI Profiles

Best for: Enterprise teams with strict compliance requirements and companies managing multiple brands with distinct voices

Cost-efficiency: High efficiency. Provides high-quality, publish-ready translations that don’t require additional investment on quality assurance.

Lokalise’s Custom AI Profiles retrieve approved examples from your past translations and translation memories to produce higher quality output with 90-95% acceptance rates. This is at par with the accuracy achieved by human translators. 

While AI orchestration works with your defined glossary and style guide, Custom AI Profiles go a level deeper to provide additional context about your brand’s terminology, tone, and writing style. It uses references from your historical translations to accurately tailor the output to your brand and specific use case. 

That means you can configure exactly how AI translation works for each domain with specific projects, content types, or language pairs.

More importantly, it’s easy to build and manage context for each profile. Instead of applying your entire translation memory for every project, create different profiles for each use case, like legal, marketing, documentation, and more. Then, curate a set of approved translation memories to use for every profile.

When should you use custom AI profiles?

Custom AI Profiles allow you to share additional context with AI for more targeted translations. 

The model uses retrieval-augmented generation to find similar entries from your translation memory or existing translations and deliver an output based on these curated references. 

Let’s understand how it works. 

To create a custom profile, choose one source language and one or more target languages. Then, select your ideal data source for this profile. Choose from two types of data sources:

• Tags: All translations saved under a specific tag(s) are fed to the AI model as context. Feel free to add one or more tags.
• Reviewed status in project: This covers all translations marked as reviewed. Select one or more projects so that all approved translations are used as context. 

For either of these data sources, Lokalise highlights the number of translated keys available for your target languages as references.

Now, select the project(s) where you want to use this profile and activate the profile to complete the setup. Once you start translating, the model uses RAG to process each string and provide targeted suggestions based on your past translations.

Choose your translation approach strategically

A successful translation project takes more than just hiring translators and linguists. You need to make strategic choices about which translation approach to go for, where to invest human expertise, and which AI models to use.

Lokalise makes these decisions easier for teams like yours. 

Our AI orchestration engine automatically routes content to the optimal AI model based on content type, language pair, and available context. The platform also scores every translation for quality, publishing high-accuracy content and flagging lower-scoring segments for human review.

Ready to build translation workflows that actually scale? Start your 14-day free trial (no credit card required) and ramp up your translation projects.