The NextGenAiAssistant – Bulk Content Generator module offers optional features for translation verification and correction, which can be enabled when launching a prompt. These tools are especially useful when translating longer texts, where the risk of context errors, language mismatches, or inconsistent phrasing is higher. Using these options is highly recommended to ensure high-quality multilingual content.
🔧 To learn how to enable verification and correction when launching a prompt, check out this configuration guide.
🔍 Supported Content Types
Currently, verification and correction are only available for translated product and category content. Support for other content types will be introduced in future updates.
🧠 Why verify translations?
Longer product descriptions are more likely to include errors or lose meaning in translation. The verification step helps to ensure that the translated content is accurate, fluent, and aligned with the original message.
The built-in translation verifiers:
- Confirm the translation is written in the correct target language
- Check grammar, structure, and overall quality
- Detect contextual inconsistencies with the original
- Improve overall translation accuracy and quality
⚠️ We recommend using the GPT-4o model for translation verification, as it provides superior understanding and language accuracy.
✏️ The correction step always uses the same AI model as the verifier, ensuring consistency between the quality assessment and the suggested changes.
⚠️ Note: Enabling translation verification and correction features may result in additional credit consumption depending on the complexity and size of the content.
🔁 Automatic Retry for Corrections
If verification fails and correction is enabled, the system will automatically attempt to fix the issue by rerunning the correction prompt.
🔁 The number of retry attempts is based on your configuration under Retrying attempts on error. Learn more in the Retry Policy guide.
🧵 Thread Management & Status Propagation
- Both verification and correction steps create separate threads, each representing a distinct operation in the content generation workflow.
- If you enable verification followed by correction, and the correction still fails verification, then all related threads (original, verification, and correction) will receive the final status:
Verification Failed after Correction - This unified status propagation helps to clearly identify failed workflows regardless of which step was the last to run.
⚙️ Feature Behavior: Correction vs. Verification Threads
- Correction threads behave like any regular prompt execution:
- Backup files are created
- Autosave is enabled
- You can continue the thread to customize the output
- Verification threads, on the other hand:
- Do not support backups or autosave
- Cannot be continued or edited manually
- Are executed silently in the background for quality assessment only
🛠️ Verification-related Statuses
| Status | Description |
|---|---|
| Completed | The content was successfully verified and is ready for use. |
| Verification Failed | Content failed verification and needs correction. |
| Verification Failed Sent for Correction | Verification failed; the content was sent back for revision. |
| Verification Failed after Correction | Content still fails after correction. All related threads reflect this status. |