Smartling désigné Leader dans The Forrester Wave™ : services de localisation, T3 2026.
Smartling nommé Leader dans la Vague™ Forrester

Updated November 20, 2026. Prices checked November 20, 2026.

The right AI translation tool depends on the job, not the brand. A quick email, a product catalog, an app release and a regulated contract each need something different, and the 12 tools below fall into four types that fit those jobs: general AI assistants, translation engines, AI-first translators and localization platforms. Each entry says what the tool is best for, what it costs, and where another tool does the job better.

If you are choosing AI translation software for a business, start with the quick picks table, then read the entries for the two or three tools that match your content.

Quick picks: which AI translation tool for which job

Match the tool to the content: engines and assistants suit one-off work, and platforms suit content that changes every week.

 

If you needCommencez parPourquoi
A quick translation of text or a document with the layout intactDeepLIts document translator takes PDF, Word and PowerPoint files and keeps the original formatting and layout, per DeepL.
A draft in a specific tone, or a translation you can steer in conversationChatGPT, Claude or GeminiYou can add instructions, a glossary and context to the prompt.
Translation built into your own product through an APIGoogle Cloud Translation or Microsoft TranslatorBoth charge by the character and include a free monthly allowance.
Files for a small team, with optional human reviewLara TranslateLara says it handles 70+ file formats and offers optional human review.
Website, product and marketing content with glossaries, translation memory and review stepsA localization platform: Smartling, Lokalise, Phrase, Crowdin or SmartcatThey connect to your CMS and code and keep approved terminology consistent across every release.

 

AI translation tools compared at a glance

Twelve tools, four types, and one starting price each. Prices are in US dollars unless marked, exclude tax, and were checked on November 20, 2026.

 

OutilTypeMeilleur pourTarif Smartling
ChatGPTAI assistantDrafting and translating with instructions in the promptFree; Go $8/month; Plus $20/month
ClaudeAI assistantDrafting and translating with instructions in the promptFree; Pro $17/month billed annually ($20 billed monthly)
GeminiAI assistant and APITranslation inside your own workflows through the APIFree API tier with limits; Gemini 3.5 Flash $1.50 input and $9.00 output per 1M tokens
DeepLMoteur de traductionText and document translationFrom €7.49/month billed annually
Google Cloud TranslationTranslation engine (API)Adding translation to a product at volumeFirst 500,000 characters a month free, then $20 per 1M characters
Microsoft TranslatorTranslation engine (API)Adding translation to a product through an APIFree tier of 2M characters a month, then pay as you go or commitment tiers
Lara TranslateAI-first translatorFiles and documents for small teamsFree (60,000 characters a month); Pro $9.99/month billed yearly
SmartlingLocalization platformEnterprise programs across website, product, marketing and support contentQuoted per contract
LokaliseLocalization platformApp and software strings, with a self-serve startFrom $149/month billed annually ($169 billed monthly)
PhraseLocalization platformEnterprise programs with many content types and teamsTeam plan $1,245/month billed annually
CrowdinLocalization platformDevelopers, growing teams and community translationFree; Pro from $75/month billed annually; Team from $450/month, annual only
Chat intelligentLocalization platform with an expert marketplaceTeams that want human experts in the same platformAdapt from $1,200 a year, billed annually

 

How we compared these AI translation tools

We looked at seven things for each tool:

  • Fit with a real workflow, from one-off text to a release that ships every week
  • Terminology control through glossaries and translation memory
  • Review options, from none to built-in human review
  • How the vendor treats your data
  • Which systems it connects to
  • How openly it publishes pricing
  • How many languages it covers

The details come from each vendor's own pricing, documentation and product pages, checked on November 20, 2026. Vendors change prices often and third-party roundups lag behind, so confirm the current price before you buy.

 

Four types of AI translation tool

The four types solve different problems, and mixing them up is the most common way to buy the wrong tool.

AI assistants (ChatGPT, Claude, Gemini) are general-purpose models you instruct in plain language. You can ask for a tone or a glossary in the prompt. They do not keep a translation memory or a review workflow unless you build one around them.

Translation engines (DeepL, Google Cloud Translation, Microsoft Translator) are built to translate. You use them in an app or through an API and pay by subscription or by the character. Terminology and review stay your job.

AI-first translators (Lara Translate) wrap AI translation in an app for documents and files, with optional human review.

Localization platforms (Smartling, Lokalise, Phrase, Crowdin, Smartcat) sit around the translation itself. They connect to your CMS and code, store approved translations and glossaries, and route content to review. In most cases they also let you choose which engine or model handles each kind of content.

 

The 12 AI translation tools, one by one

Every entry follows the same pattern: what the tool is best for, what it costs, and where another tool is the better choice.

 

Assistants IA

 

ChatGPT

Best for: Drafting, adapting and translating text when you can give instructions in the prompt.

ChatGPT translates what you paste or upload, and you can tell it the audience, the tone and the terms to keep. That makes it useful for first drafts and for adapting copy, not only converting it. It has no translation memory or approval workflow built for localization, so keeping terms consistent across a team is up to you. On the individual plans, OpenAI's pricing page lists content as used to train its models, with an opt-out available; check the terms for Business and Enterprise.

Price: Go is $8 a month and Plus is $20 a month; Pro starts at $100 a month. Business is $20 per user a month billed annually ($25 billed monthly) for teams of 2 or more. Enterprise is quoted. Per OpenAI's pricing documentation and Business pricing page.

Where another tool wins: A translation engine or platform fits better when you need the same glossary applied every time, review steps, or a connection to your CMS.

 

Claude

Best for: Drafting, adapting and translating text when you can give instructions in the prompt.

Claude works the same way: you describe the audience, the tone and the terms to keep, and it translates or adapts the text. Like the other assistants, it does not come with a translation memory or review workflow, so consistency across many files depends on the prompts you reuse.

Price: Free; Pro $17/month billed annually ($20 billed monthly); Team $20 per seat per month billed annually ($25 monthly). Through the API, Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, per Claude's pricing page.

Where another tool wins: A platform fits better when you need approved terminology, an audit trail or an integration with your content systems.

 

Gemini

Best for: Translation inside your own workflows through the Gemini API.

Developers can call Gemini to translate text and pay per token. Google describes Gemini 3.1 Flash-Lite as suited to translation, and batch requests cost 50% less. The free tier has limits, and Google's pricing page notes that free-tier content may be used to improve its products, so keep confidential text off it.

Price: Free API tier with limits. Gemini 3.5 Flash is $1.50 per million input tokens and $9.00 per million output tokens; Gemini 3.1 Flash-Lite is $0.25 and $1.50.

Where another tool wins: A platform fits better when non-developers need to run translation, or when glossaries and review matter more than per-token cost.

 

Translation engines

 

DeepL

Best for: Text and document translation.

DeepL's document translator takes PDF, Word and PowerPoint files and keeps the original layout. It supports 100+ languages, per DeepL, and the paid plans add glossaries and style rules. Business adds translation memory and SCIM, and DeepL lists SOC 2 Type 2 and ISO 27001. DeepL also says its translation is available inside ChatGPT, Claude and Microsoft Copilot through MCP.

Price: In euros, VAT included, per DeepL's pricing page: Individual from €7.49 a month billed annually (300,000 characters); Team €24.99 per user per month (1,000,000 characters per user); Business €49.99 per user per month (unlimited characters under fair use). Enterprise is custom.

Where another tool wins: A localization platform fits better when translation has to flow through your CMS or code and a review step.

 

Google Cloud Translation

Best for: Adding translation to a product or workflow at volume through an API.

Google Cloud Translation is the API for building translation into your own software. It is an engine, so someone has to build the workflow around it. For text inside images, Google's photo translation tutorial pairs Cloud Vision, which detects the text, with Cloud Translation, which translates it, so image translation is a two-step build.

Price: The first 500,000 characters a month are free, then $20 per million characters.

Where another tool wins: A platform fits better when marketers and translators need to work without engineering help.

 

Microsoft Translator

Best for: Adding translation to a product or workflow through an API.

Microsoft Translator is a cloud service for text and document translation that you call from your own systems. As with Google, the engine translates and your team builds the process around it.

Price: A free tier of 2 million characters a month, then pay-as-you-go pricing and commitment tiers for high volumes. Microsoft's pricing page lists the current rates.

Where another tool wins: A platform fits better when you need glossaries, approvals and connectors to your content systems out of the box.

 

AI-first translators

 

Lara Translate

Best for: Documents and files for individuals and small teams.

Lara says it supports 200+ languages and 70+ file formats, and it offers optional human review. Its Team plan covers a whole team for one price, with unlimited users and a shared character allowance.

Price: Per Lara's pricing page: Free ($0, 60,000 characters a month); Pro $9.99 a month billed yearly (500,000 characters); Team $29.99 a month for the whole team billed yearly (1.5M shared characters). Enterprise is custom.

Where another tool wins: A platform fits better when translation has to connect to your CMS and code and move through approvals across several teams.

 

Localization platforms

 

Smartling

Best for: Enterprise teams running AI translation across website, product, marketing and support content in one governed program.

Smartling lets you choose from 20+ LLMs and machine translation engines, or bring your own provider keys, and pulls your glossary and translation memory into each LLM translation. Its Language Quality Estimation Agent, part of the paid AI Toolkit add-on, predicts quality for each machine-translated string so strings skip or go to human review by rule, and its LQA tools score output with MQM. Human translators work in the same platform: Smartling lists 4,000+ linguists and also lets you bring your own. Customizable workflows, linguistic asset management and an LQA suite apply across every content type. Smartling says its AI Hub translates into 450+ languages, and Forrester named it a Leader in The Forrester Wave: Localization Services, Q3 2026. See how the AI Hub works.

Price: Enterprise pricing is quoted per contract.

Where another tool wins: Lokalise fits better if you want to start without a sales call or your content is mostly app strings that need over-the-air mobile updates. It also fits if your security team requires data stored only in the EU: Smartling hosts data on AWS in the US only.

 

Lokalise

Best for: App and software strings, with a self-serve start.

Lokalise is built around app and software strings, with iOS, Android and Flutter SDKs that push string updates over the air. It offers a 14-day trial with no credit card, includes AI translation words in every plan and scores each string, and, per Lokalise, routes strings automatically among 5+ LLMs. Its newer Vantage workspace gives marketing, legal and support teams a no-code way to request and manage translation. Lokalise states it does not store customer data outside the EU.

Price: Per Lokalise's pricing page: Explorer from $149 a month billed annually ($169 billed monthly); Growth $379 a month billed annually ($439 billed monthly); Advanced $1,049 a month billed annually; Enterprise by quote. Lokalise meters processed words, advanced seats, integrations and automations.

Where another tool wins: Smartling fits better when website, marketing and support content, human translators and a website proxy need to run in one governed program. Smartling vs Lokalise.

 

Phrase

Best for: Organizations that want translation workflows, machine translation profiles and many integrations in one platform.

Phrase's Team plan includes unlimited TMS seats, customizable translation workflows, integrations to Figma, Slack, WordPress and more, and glossaries and profiles for its AI-powered machine translation. Business adds custom metadata and branching for products and integrations such as Drupal, Zendesk and Kontent.ai. Every plan comes with a 14-day free trial and 24/7 support, per Phrase.

Price: Per Phrase's pricing page: Team $1,245 a month billed annually; Business and Enterprise are quoted. Freelancer and agency plans are listed separately.

Where another tool wins: Lokalise or Crowdin fit better for a small team that wants to start below $1,245 a month. More platform comparisons are on Smartling's compare hub.

 

Crowdin

Best for: Developers, growing teams and community translation.

Crowdin's pricing is built around hosted words, and its Team plan includes 5 managers, a termbase, style guides, customizable workflows with source text review, quality checks, Crowdin AI and AI pipeline auto-translation. Crowdin says it connects to 700+ apps and integrations. Business adds SSO through SAML, an audit log and a dedicated account manager as an add-on.

Price: Per Crowdin's pricing page, excluding tax: Free ($0, 40K hosted words); Pro from $75 a month billed annually ($89 billed monthly); Team from $450 a month, annual billing only ($5,400 a year, 500K hosted words); Business is quoted.

Where another tool wins: Enterprise platforms such as Smartling or Phrase fit better when a central team governs many content types and vendors. Smartling lists its other platform comparisons on its compare page.

 

Chat intelligent

Best for: Teams that want AI translation and human experts in one platform.

Smartcat says its AI coworkers translate, review and update content in 280+ languages. It offers a marketplace of 500K+ human experts, included on its higher plans and an add-on on the entry plan. Translation memory matches are free on every plan. The entry plan connects to Google Drive, Google Sheets, a Chrome extension and OneDrive; CMS and developer-pipeline integrations start on the third plan, Anticipate.

Price: Per Smartcat's pricing page: Adapt from $1,200 a year billed annually (100K Smartwords, unlimited users); Accelerate, Anticipate and Autonomous are custom. Smartwords are an annual allowance, and the 15-day trial includes API access.

Where another tool wins: If your content lives in a CMS or code repository, check the plan first, because Smartcat's connectors for them start on its third tier.

 

Which AI translation tool fits which team

Start from who does the work and how often the content changes.

 

Your teamCommencez parPourquoi
One person translating the occasional documentLara Translate or DeepLBoth are apps with document translation; Lara has a free plan, and DeepL's Individual plan starts at €7.49 a month.
Developers adding translation to a productGoogle Cloud Translation, Microsoft Translator or the Gemini APIThey are priced by the character or the token and called from your own code.
A product team shipping app strings every releaseLokalise or CrowdinBoth are built around app and software strings, and both publish self-serve prices. Phrase fits when many teams share the program.
A marketing team with a website, campaigns and a help centerSmartling, Phrase or SmartcatThese connect to your CMS. Smartling also offers a website translation proxy and connectors for AEM, Contentful, HubSpot, Shopify, Sitecore, Webflow and WordPress.
A security-reviewed enterprise programSmartling or Lokalise, among othersCompare certifications on the security pages. Smartling lists SOC 2, ISO 27001 and 42001, HIPAA, HITRUST e1, PCI Level 1 and GDPR; Lokalise lists SOC 2 Type II, ISO 27001 and 27017, and GDPR.

 

For more depth, read our guide to machine translation software, our guide to the best translation management software, and our vendor comparison pages.

 

Where AI translation still falls short

AI translation is fast and cheap enough to use widely, but two problems come up across every type of tool.

Quality and consistency vary. Output can be strong in one language pair or content type and weaker in another, and fluent text can still be wrong. That is why several platforms score each string for quality: Smartling's Language Quality Estimation Agent and Lokalise's AI scoring both decide which strings need a person to look at them. Terminology drifts too: without a glossary or translation memory, the same product name can come back several different ways. Assistants and engines leave that to you, and platforms keep approved terms and apply them every time.

Your content may not stay private. ChatGPT's pricing page lists content as used to train models on its individual plans, with an opt-out available. Google notes that content sent on the free Gemini API tier may be used to improve its products. Smartcat says it never uses customer content to train public AI models. Read the data terms before you translate contracts, HR files or unreleased product copy.

For legal, regulated and high-visibility pages, add a human review step, whichever tool does the first pass.

 

How AI translation tools are priced

Five pricing units cover all 12 tools, and they are not comparable one for one, so work out your own volume before you compare prices.

 

Priced byWhat you pay forExemples
CharacterThe text you sendGoogle Cloud Translation: first 500,000 characters a month free, then $20 per 1M. Microsoft Translator: free tier of 2M characters a month.
TokenText in and text out of a language modelGemini 3.5 Flash: $1.50 per 1M input, $9.00 per 1M output. Claude Sonnet 5.5 API: $2 and $10 per 1M.
UserA subscription for each personClaude Team $20 per seat a month billed annually; DeepL Team €24.99 per user a month; ChatGPT's paid plans per user a month.
Word allowanceA pool of words you draw downCrowdin hosted words (Team: 500K); Smartcat Smartwords (Adapt: 100K a year); Lokalise processed words.
ContractA quote based on volume, languages and servicesSmartling; Phrase Business and Enterprise; Smartcat Accelerate and above.

 

Ask each vendor what counts against the allowance, including repeated sentences and added languages. Smartcat says translation memory matches are free on every plan.

 

Planning a wider rollout? Our free ebook, Navigating the shift: Why, when, and how to adopt AI translation, covers where to start.

FAQ

What is the best AI translation tool for business?
It depends on the content. For the occasional document, DeepL or Lara Translate is enough. For website, product and marketing content that changes every week, use a localization platform such as Smartling, Lokalise, Phrase, Crowdin or Smartcat, because they add glossaries, translation memory and review steps.
Is ChatGPT good for translation?
ChatGPT translates and adapts text, and you can set the tone and the terms to keep in the prompt. It does not include a translation memory or an approval workflow, and on its individual plans OpenAI lists content as used to train its models, with an opt-out available. For ongoing business content, a platform is a safer base.
What is the difference between DeepL and Google Cloud Translation?
Both are translation engines. DeepL offers an app, a document translator and per-user plans from €7.49 a month, and it supports 100+ languages. Google Cloud Translation is an API priced by the character, with the first 500,000 characters a month free.
Can AI translate text inside images?
Yes, but it is usually two steps: software reads the text out of the image, then translates it. Google documents this pattern with Cloud Vision plus Cloud Translation. You can also upload an image to a general AI assistant and ask it to translate the text. Either way, someone still has to put the translated text back into the design, and that is where a localization platform with visual context helps.
Can you publish AI translations without human review?
Sometimes. Low-risk content such as internal documents or help articles can go out with automated quality checks, while legal, regulated and customer-facing pages should get a human review. Several platforms score each string so reviewers only see what needs attention.
How much do AI translation tools cost?
From free to a quoted contract. Lara Translate has a free plan, Claude Pro is $17 a month billed annually, DeepL starts at €7.49 a month, Lokalise at $149 a month billed annually, Crowdin Team at $450 a month, Phrase Team at $1,245 a month and Smartcat at $1,200 a year. Smartling prices per contract.
Do AI translation tools keep my content private?
It depends on the vendor and the plan. ChatGPT's individual plans list content as used for training with an opt-out, and Google notes that free Gemini API content may be used to improve its products. Smartcat says it never uses customer content to train public AI models. Check each vendor's data terms before you translate confidential text.
What is the difference between a translation engine and a localization platform?
An engine translates text. A platform manages the work around it: connectors to your CMS and code, glossaries, translation memory, review steps and a choice of engines.
What are alternatives to Lara Translate for business?
DeepL is an alternative for document translation. If you need glossaries, translation memory and approvals across a team, look at a platform such as Smartling, Smartcat or Phrase. Lara says it offers optional human review.
Which AI translation tools have a free plan?
ChatGPT, Claude, Lara Translate and Crowdin (for public projects) have free plans. Google Cloud Translation includes 500,000 free characters a month, Microsoft Translator a free tier of 2 million characters a month, and the Gemini API a free tier with limits.
Can AI translation keep the formatting of a document?
Several tools can. DeepL says its document translator keeps the original layout of PDF, Word and PowerPoint files, Lara says it supports 70+ file formats, and Smartcat says it handles 80+ file formats.

Reagan White

Expert en localisation
Reagan White est une experte en localisation qui aide les marques internationales à rationaliser les flux de traduction et à développer des contenus multilingues. Avec une formation en technologie de la traduction et en stratégie de contenu international, elle écrit sur l'automatisation de la localisation, la traduction IA et les meilleures pratiques pour construire des opérations mondiales efficaces.

Pourquoi attendre pour traduire de manière plus intelligente ?

Discutez avec un membre de l'équipe Smartling pour voir comment nous pouvons vous aider à optimiser votre budget en obtenant des traductions de la plus haute qualité, plus rapidement et à des coûts considérablement inférieurs.
Cta-Card-Side-Image