Machine translation is the automated conversion of source text into target text by a software system, without direct human translation during the initial generation. In localization production, it can be used as a draft source, as raw input for post-editing, or as a gisting tool, but it is evaluated separately from human translation because quality varies by content type, language pair, and model.
Machine translation appears in translation workflows when clients supply MT output for post-editing, when teams use MT engines through CAT tools or APIs, and when reviewers compare raw MT quality for language pairs. It is documented in workflow briefs, quality plans, and cost estimates because it changes productivity, pricing, and review effort.
Teams sometimes treat raw machine translation as final quality or reuse MT suggestions for sensitive, creative, or legal content without expert review, leading to accuracy errors, tone problems, or confidentiality risks. In regulated or brand-critical contexts, that assumption can create serious liability, damage trust, and require costly corrective revision after publication.
For creative and consumer content the failure mode is flatness, not inaccuracy — and flatness passes an accuracy check. Terms here also carry technical constraints (reading speed, character budgets) that are part of the brief.
Angel Translation locks terms like this into a project terminology base before expert review begins, so the same source term resolves to the same target term across every document in a submission — and stays consistent in the next revision. See the 8-step AI + expert workflow.
Send us the files. Acknowledged in 4 hours, quoted in 24.
Request a Quote →Content Globalization solutions →