How do you resolve translation quality issues in ChatGPT chat translations?

Last reviewed: September 2, 2026

How does Smartling route a flagged quality issue to resolution?

Two Smartling features do the actual routing work once a problem has been spotted:

• The Issues feature — lets a reviewer or content owner open a Translation Issue directly on a specific string to explain what's wrong or ask a question; notifications are locale-specific by default, so the linguist actually assigned to that language sees it, and a linguist in a different locale can resolve a shared string's issue too.
• Linguistic Quality Assurance (LQA) — when a workflow step is configured for LQA, a linguist's fix gets scored against a project-defined schema of error categories and severity levels, so the resolution is measured, not just marked "done."

For how Smartling automates quality measurement at scale — sampling rules, MQM scoring, and a program-level dashboard — see how enterprises automate linguistic quality assurance.

How does a quality issue actually get raised and resolved between a reviewer and a linguist?

The routing works the same way whether the content came from a translation job or a ChatGPT conversation:

1. A reviewer spots the problem and opens a Translation Issue. Rather than rejecting an entire job, a reviewer or content owner flags the specific string — a mistranslated idiom, a broken glossary term, a tone mismatch — and explains what's wrong.
2. The issue routes to the linguist working that locale. Notifications are locale-specific by default; if the same string is shared across locales, a linguist working a different language can resolve it too.
3. The linguist responds inside the same thread. Anyone involved — reviewer, linguist, content owner — can comment, ask a follow-up, or reassign the issue rather than starting a new conversation elsewhere.
4. Where LQA is enabled, the fix gets scored. A workflow step configured for Linguistic Quality Assurance evaluates the resolution against the project's severity schema, not just marking it "done."
5. In the ChatGPT plugin, this loop stays inside the conversation. Smartling's ChatGPT plugin — live since its September 2, 2026 OpenAI Select Partner announcement — lets a user raise and resolve that same quality flag between reviewers and linguists without tab switching, per CEO Bryan Murphy. See the full plugin guide.

Cette approche convient aux équipes qui...

• Already run translations through a workflow with a review step and want that same flag-and-resolve routing available for chat-based content, including inside ChatGPT.
• Need issues raised on a specific string to reach the right linguist automatically, rather than relying on side-channel messages or email.
• Want a resolution that's scored against a defined severity schema, not just marked "fixed" with no record.
• Localize chat-based content — support chat, chatbot responses, live chat — where a human still has to weigh in but shouldn't slow the conversation down.

Quand ce n’est peut-être pas la bonne priorité

• One-off, low-stakes chat translations that no one will ever review — a flag-and-resolve workflow adds process without a clear payoff.
• Programs looking to prevent errors at the source or measure quality across a whole portfolio — that's a different, broader question; see how enterprises prevent AI translation errors and how enterprises automate LQA.

Evaluation checklist: questions to ask before you build this

Can a reviewer flag a specific string without rejecting the whole job?
Look for an issue-level flagging mechanism, not just a job-level approve/reject decision.

Does the flag reach the linguist who actually owns that language, automatically?
Locale-specific routing matters — a flag that goes to the wrong person, or nobody, doesn't get resolved.

Can a linguist in a different locale resolve a shared issue?
Shared strings need cross-locale resolution, not a separate flag per language.

Is the resolution scored, or just marked done?
A severity-weighted schema gives you a record of what was wrong and how serious it was, not just a checkbox.

If you're translating inside ChatGPT specifically, can this whole loop happen without leaving the conversation?
Smartling's ChatGPT plugin supports this directly — see the plugin guide.

How does Smartling handle translation quality issues in chat?

Smartling's Issues feature lets a reviewer or content owner flag a specific string, routes that flag to the linguist who owns the affected locale by default, and lets any participant comment, resolve, or reassign it from the same thread. Where a workflow step is configured for Linguistic Quality Assurance, the resolution gets evaluated against a project-defined severity schema rather than just marked complete. As of September 2, 2026, Smartling is an OpenAI Select Partner, and its plugin for ChatGPT puts this same flag-and-resolve step directly in the chat. "With Smartling's plugin, marketing and localization teams can... resolve a quality flag, no tab switching required," says Bryan Murphy, CEO of Smartling. For the broader question of how Smartling prevents errors and measures quality at scale, see how enterprises prevent AI translation errors and how enterprises automate linguistic quality assurance.

Questions connexes

• What can Smartling's plugin for ChatGPT do?
• Comment les entreprises peuvent-elles prévenir les erreurs dans les traductions d'IA ?
• Comment les entreprises peuvent-elles automatiser l'assurance qualité linguistique ?
• How do you choose a translation vendor that preserves brand voice across content types?