This article describes how to set up your integration with OpenAI in POEditor and how to use models like GPT in your localization projects to translate, adjust and evaluate translations.

AI translation and evaluation features are available on any POEditor subscription plan, including the Free plan.

How to set up the OpenAI integration for localization

To set up the OpenAI integration, go to Integrations and select OpenAI.

Enter your OpenAI API key, confirm that you understand AI features are experimental, then select the model you want to use and save your settings.

You can also set OpenAI as your default AI provider. This default provider will be used whenever you run AI translation, AI adjustment, or Quality Evaluation, unless another provider is selected.

Integrations are set up at account level. If you're an admin or contributor in a POEditor localization project, you can set up your own AI integration or ask the project owner to share their integration with you as described in the next section.

How to share your OpenAI integration with contributors

Your OpenAI integration can also be shared with contributors working on your localization projects.

In your Account Settings, enable the Contributors Can Use Your AI Integration option. Contributors will then be able to use your OpenAI integration without adding their own AI API key.

How to translate strings with OpenAI

Once your OpenAI integration is configured, you can start translating strings from any language page.

For individual strings, click an empty translation field and select Translate with AI.

To translate multiple strings at once, use the AI menu at the top of the Language page.

Choose the source language and the target language(s), then optionally use Advanced Settings to limit AI translation to strings with specific tags. Only strings without translations from your selection will be translated.

AI post-editing with OpenAI (adjustment prompts)

OpenAI can be used to adjust existing translations, whether they were generated with AI, imported, or added manually.

After an AI translation has been generated, you can request a different version of the translation based on your requirements. The available adjustment prompts include Formal, Simplify, Shorter and Rephrase. You can also add your own adjustment prompts.

Existing strings can also be modified using the option Edit with AI.

When fixing issues detected by Quality Evaluation, you can use the QE Fix adjustment prompt for existing translations.

AI quality evaluation with OpenAI

Quality Evaluation is another AI-powered feature available through your OpenAI integration.

It scores each translation out of 100 using the MQM (Multidimensional Quality Metrics) framework, and breaks down any problems found by category and severity.

To use Quality Evaluation, the project must have a Default Reference Language set, or another project language available to use as the reference language.

If your project has a glossary assigned, the relevant glossary terms and definitions are automatically included in the evaluation.

To run Quality Evaluation, open a language page and choose a scope:

• Unscored translations only - runs the evaluation only on translations that don't have a score yet.
• All translations - re-evaluates everything, including translations that were previously scored.

Scores appear next to each translation on hover.

Click any score to see the full breakdown of flagged issues.

From there, click Fix with AI to generate a corrected version. Review the suggestion in the AI Translation panel and click Use it to apply it, or generate another variation using the available adjustment prompts.

How to customize AI prompts

The default AI prompts are optimized for performance and cost efficiency, but you can customize them to better suit your workflow.

In your OpenAI integration settings, click Add custom prompt to edit the default translation prompts (Translate, Translate Plural and Translate Bulk), the built-in adjustment prompts (Formal, Simplify, Shorter, Rephrase and QE Fix), or add your own adjustment prompts.

Translation prompts are used for AI translation, while adjustment prompts are used for editing existing translations, including QE Fix for Quality Evaluation results.

When customizing prompts, make sure to keep the required placeholders. All adjustment prompts must include the placeholder {target_translation}.

If your project uses plural forms, you'll need to customize both the Translate and Translate Plural prompts.

Custom models

If you've trained a custom AI model with OpenAI, you can use it in POEditor instead of one of the default options.

Once you've connected your OpenAI integration, your custom model will appear in the model selector, where you can select it for use.

Workflows: how to add an AI translation step

You can add a Translate with AI step to any Workflow to automatically translate strings using your default AI provider.

Some workflow templates already include this step, or you can also add it to your own Custom workflows.

Select the source language, where the source strings should be taken from (Project Terms or a project language), and the target languages. If the project has a Default Reference Language, it will be selected automatically.

You can translate languages inherited from the previous step, a custom selection, or all project languages. Optionally, use Advanced options to filter the strings sent for translation by tag.

Using the official POEditor MCP server

If you use an MCP-compatible AI assistant, you can connect it to POEditor using the official POEditor MCP server.

This lets you manage projects, translations, exports, QA checks and other localization tasks directly from your AI assistant, without writing API scripts or switching to the POEditor interface.