AI post-editing (AIPE) services for high-volume content
Our AI post-editing (AIPE) services give high-volume programs more than a corrected file. You get terminology that holds from batch to batch, a native post-editor whose time goes to meaning and voice, and evidence with every delivery that you can show your stakeholders. The AI is calibrated on your glossary and rules before it translates, agents check the text before and after the linguist, and an automated quality gate runs on every delivered file. If a file fails the gate, it does not ship.
What AIPE adds to MTPE
Our AI post-editing (AIPE) builds a managed process around the human edit. In typical machine translation post-editing (MTPE), a post-editor corrects the machine output and the edited file is what you receive, though workflows vary by provider. Your glossary, translation memory (TM), and published terms are prepared before the AI translates. Agents check the text before and after the native post-editor, the rules that can be tested reliably run as automatic checks, and an automated quality gate checks every delivered file. The post-editor’s time goes to meaning and voice, every delivery carries the evidence, and your reviewers’ edits feed into the next batch.
| Row | Typical MTPE | Locstars AIPE |
|---|---|---|
| The AI output | Machine output, with engine customization that varies by provider | AI translation calibrated on your glossary, TM, and style rules for every job |
| Terminology | Often looked up by the post-editor during the edit | Researched from your own glossary, TM, and published content before translation, with a source for every term |
| Checks around the linguist | Often, the post-editor is the main check | A native-reader agent and independent quality assurance (QA) checks before and after the edit. At AI + Edit + Review, also a red team and a judge agent |
| Your rules | Often a style guide the post-editor has to remember | Rules that can be tested reliably run as automatic checks on every file. Linguists judge meaning and tone |
| Before delivery | The edited file is sent | An automated quality gate checks every delivered file. A file that fails does not ship |
| What you receive | Usually the translated file | Your files, the quality-gate result, an issue log, glossary changes, and a query log with every delivery, plus quality reports: sampled on an agreed schedule at AI + Edit, full with every delivery at AI + Edit + Review |
| Your reviewers’ edits | Usually corrected file by file | Compared with what we delivered. Each confirmed lesson becomes a glossary entry, a rule, or an automatic check before your next batch |
AIPE runs at AI + Edit and AI + Edit + Review, where a native post-editor edits every segment. If you buy MTPE today, AI + Edit is the level to compare it with. A sample on your own file lets you make that comparison on your content.
What you send
- Files in your format, such as DOCX, XLSX, CSV, XML, JSON, XLIFF, HTML, or PO, including exports from your translation management system (TMS) or content management system (CMS)
- Or access to a job in your TMS, including Phrase, XTM, Smartling, Crowdin, and Lokalise
- Your glossary, translation memory (TM), and style guide, if you have them
- Content you have already published in the target language, if any
- The content types in the batch, so each one gets its service level
- Any instructions or reference material
No glossary or style guide yet? We create them with you before production starts.
What happens to your files
Every step has a named owner, so you always know what an agent did and what a linguist decided.
- Intake and scoping (all levels). Agents measure your files and find what changed since your last batch. A project lead confirms the scope and the service level for each content type, and any open questions come to you before translation starts.
- Terminology with sources (all levels). Agents propose terms, each with its source, starting with your glossary, your TM, and your published content. A form seen once is a candidate, not a rule. You approve the initial glossary and every new key term (product, feature, and brand names, and legal or regulated terms). Other new terms are approved by your linguist and added to your glossary, which we share whenever you ask.
- Calibrated AI translation (all levels). Agents translate against your glossary, TM, and rules, then run automatic checks on the result.
- Native read and linguist edit (all levels, at different depths). A native-reader agent sees only the target text and flags anything that reads like a translation. The linguist then works at your level’s depth.
- Independent checks (AI + Edit and AI + Edit + Review). Quality assurance (QA) agents check accuracy, grammar, style, and terminology, each blind to the others’ results.
- Debate and decision (AI + Edit + Review). Advocate and devil’s-advocate agents argue each finding, a judge agent recommends, and the reviewer decides and records the reason.
- Quality gate and report (all levels). An automated gate checks every delivered file before it reaches you.
Every job runs steps 1–4 and 7. Steps 5 and 6 run at the higher service levels, as named in your quote.
A finding counts only when it names a concrete defect and points to a reference: the source, your glossary, a rule, or your instructions. Preferences are logged and never block delivery.
AI post-editing at AI + Edit and AI + Edit + Review
At AI + Edit, our default level, a native post-editor edits every segment of the AI output. This is full post-editing: the goal is text that reads as if it were written in the target language. After the edit, QA agents check the result, and a reviewer runs a sampled quality check.
At AI + Edit + Review, a second linguist then reviews every segment and does a final read of the delivered file. A red team of agents tries to make the file fail, and it must back every claim with evidence. A script applies the agreed changes, and a regression check re-reads everything that changed.
Each linguist is a native speaker of the target language and a subject specialist.
Example (synthetic text, English to French)
| Row | Text |
|---|---|
| Source | Your order ships within two business days. |
| AI output | Votre commande expédie sous deux jours ouvrés. |
| Post-edited | Votre commande est expédiée sous deux jours ouvrés. |
| Reason | Fluency, minor: “expédier” takes a direct object, so the sentence needs the passive. The order is shipped; it does not ship anything. |
AI translation with a linguist check at AI + Check
Some content needs to be accurate and clear, and nobody will judge its style: internal knowledge, archives, user reviews, and long-tail catalog attributes. Full post-editing there spends linguist time where it adds little. That is what AI + Check is for: AI translation with a linguist check.
Alongside the native-reader agent, an accuracy agent compares source and target and flags meaning risks. A native linguist checks every flagged segment plus a random sample, fixes the errors, and settles every flag, so the linguist’s time goes where the risk is. AI + Check is not post-editing. It never runs on regulated text, UI strings, checkout and payment text, or brand copy.
Your file, and the proof of what was checked
You don’t have to read the target language to know what happened to your file. Every delivery includes the quality-gate result and an issue log. At AI + Edit, sampled quality reports follow the schedule agreed for your program and can cover several small releases; at AI + Edit + Review, each delivery gets a full report.
- Files in your format, with tags and placeholders preserved
- The quality-gate result and an issue log
- Quality reports: sampled at AI + Edit, on the schedule agreed for your program, or a full report with every delivery at AI + Edit + Review
- Glossary changes and a query log at AI + Edit and AI + Edit + Review
- Your TM, updated, with a Translation Memory eXchange (TMX) file on request
The quality report classifies issues as critical, major, or minor, across accuracy, fluency, terminology, style, locale, tags and format, and client requirements.
If you find a language issue within the agreed scope, we correct it free of charge. Every edit your reviewer makes is compared with what we delivered. Each confirmed lesson becomes a glossary entry, a rule or an automatic check before your next batch.
Same-day delivery is possible for suitable volumes. We confirm timing in the quote.
Choosing no AI for some files
Some files may need to stay out of AI altogether. Tell us which, and they are translated by native linguists without machine translation or generative AI, using your translation memory and glossary. You can mix routes in one program: AI + Edit for your help center, for example, and no AI for your contract templates. Regulated and high-risk content runs at AI + Edit + Review or on the no-AI route.
Content that does go through AI is processed only through business AI services whose terms bar training on it.
Questions about post-editing
What is the difference between a linguist check and full post-editing?
At AI + Check, a native linguist checks the segments our agents flag, plus a random sample, and fixes what is wrong. At AI + Edit, a native post-editor edits every segment, so the text reads as if it were written in the target language.
Is AIPE the same as MTPE?
No. Both include a human edit of machine output. In typical MTPE, the edited file is the deliverable. Our AIPE builds a managed process around that edit: your glossary, TM, and style rules are prepared before the AI translates, agents check the text before and after the native post-editor, the rules that can be tested reliably run as automatic checks, and an automated quality gate checks every delivered file. Every delivery comes with the quality-gate result and an issue log.
Which level fits our content?
It depends on risk. Content that readers only need to understand can run at AI + Check. Help-center articles, UI strings, and website pages usually run at AI + Edit. Campaigns and regulated and high-risk content run at AI + Edit + Review, or without AI. See which level fits which content.
Do you use our translation memory and glossary?
Yes. Terms come from your glossary, TM, and published content first, and the AI translates against them. We turn the rules that can be tested reliably into automatic checks. Linguists judge meaning, tone, and anything else that needs judgment.
Can you work inside our TMS?
Yes. We work inside your platform, including Phrase, XTM, Smartling, Crowdin, and Lokalise, or from the exports you send. Files come back in your format, with tags and placeholders preserved.
What if we find an error after delivery?
If you find a language issue within the agreed scope, we correct it free of charge. Your reviewer’s edits are also compared with what we delivered, and each confirmed lesson goes into your glossary or rules before the next batch.
Is our content used to train AI models?
No. Your content is processed only through business AI services whose terms bar training on it, and you can choose no AI for any content type. See how we handle your content.