A tender asks whether your organization provides “human translation” or is certified to ISO 18587. Treating those as interchangeable can produce an inaccurate response, an unsuitable process design, or audit evidence that does not support the claim. The ISO 18587 vs human translation question is not a choice between quality and technology. It is a question of service scope, process control, competence, and the standard against which conformity is assessed.

For language service providers, the distinction matters commercially. Buyers increasingly ask how machine translation is used, what level of human intervention is applied, and whether the resulting service is governed by recognized requirements. A defensible answer requires more than stating that linguists review output. It requires a documented process that matches the service being delivered.

ISO 18587 vs Human Translation: The Core Difference

ISO 18587 specifies requirements for the full post-editing of machine translation output. It is designed for language service providers that deliver machine-translation post-editing, commonly called MTPE, as a professional service. The standard addresses the resources, competencies, processes, and quality controls needed when a human post-editor brings machine-translated content to an agreed final quality level.

Human translation, by contrast, describes a service method rather than a single ISO standard. In a conventional translation workflow, a qualified translator creates the target text from the source text. Where a provider seeks a recognized framework for this work, ISO 17100 is generally the relevant standard. It defines requirements for core translation processes, including qualified personnel, project preparation, translation, revision, final verification, and handling of client information.

This distinction is decisive: ISO 18587 does not certify that machine translation alone is acceptable, and it does not replace the controls expected for human translation services. It concerns a defined post-editing process. ISO 17100 concerns translation services where the translation is produced by a human translator, with revision carried out by a second qualified person when the standard’s requirements apply.

Why “Human in the Loop” Is Not Enough

Many procurement documents use broad wording such as “human-checked machine translation.” That wording is not sufficiently precise for an ISO conformity claim. A human may review output lightly for usability, correct selected errors, or fully post-edit it to a level comparable with human translation. Those activities have different objectives, risks, time allocations, and controls.

ISO 18587 is concerned with full post-editing. The post-editor is expected to ensure that the final content is accurate, complete, understandable, and appropriate for its intended purpose. The process must address linguistic correctness, terminology, style requirements, formatting, and the correction of errors introduced by the machine translation system. The result is not simply “acceptable MT output.” It is a controlled service delivered against agreed specifications.

A basic quality check by a bilingual employee, automated quality assurance tool, or project manager does not automatically meet the requirements of full post-editing. During an assessment, an organization should be able to demonstrate what was ordered, what level of post-editing was agreed, who performed the work, which checks were completed, and how the final deliverable was approved.

Process Differences That Affect Compliance

The operational difference starts before the first segment is processed. For human translation under ISO 17100, the provider must determine client requirements and assign competent personnel for translation and revision. The translated text is created by the translator, then revised by a separate qualified reviser unless an authorized and documented exception applies within the organization’s system.

For ISO 18587, the provider must additionally control the machine translation component. This includes determining whether machine translation is suitable for the content and intended use, identifying the post-editing requirements, and ensuring that the selected technology does not compromise confidentiality, security, or client instructions. The organization must also ensure that post-editors are qualified for both translation and post-editing work.

The distinction affects the production record. A human translation project should show evidence of translator and reviser assignment, revision activity, final verification, and release. An MTPE project should show the applicable post-editing specification, machine translation environment or engine controls, post-editor assignment, quality procedures, and acceptance criteria. If the organization offers both services, its procedures should make the routing decision clear.

Competence Must Match the Assigned Role

In both service models, competence cannot be assumed merely because an individual is bilingual. ISO-based assessment examines whether personnel meet defined competence criteria and whether the provider maintains records supporting those criteria.

For human translation, competence records commonly cover translation qualifications, documented professional experience, subject-matter knowledge, language competence, and ongoing professional development. For ISO 18587 services, post-editors need comparable linguistic and domain competence, together with demonstrated capability in full post-editing. They must understand the characteristic failure modes of machine translation, including omissions, incorrect terminology, misleading fluency, inconsistent phrasing, and formatting defects.

The organization should also define who can make service decisions. A project manager should not downgrade a full post-editing requirement to a lighter review simply to meet a deadline or price target. Changes to scope need approval, communication, and traceability.

Quality Expectations Are Similar, but Risk Controls Differ

Clients often assume that full post-editing and human translation will always produce identical results. That expectation may be appropriate for certain content, but it should be confirmed in the service agreement rather than presumed. The suitability of MTPE depends on content type, language pair, terminology maturity, machine translation performance, intended audience, legal or regulatory exposure, and client tolerance for variation.

For high-risk content, such as regulated medical information, legal instruments, safety documentation, or material used in formal proceedings, the organization should conduct a documented suitability review. Machine translation may be unsuitable, or additional controls may be necessary. A standard does not remove the provider’s responsibility to evaluate risk and communicate limitations.

Quality assurance also differs in emphasis. Human translation quality control focuses on the accuracy and completeness of the translator’s work and the effectiveness of revision. MTPE quality control must additionally detect errors that are persuasive on the surface but wrong in meaning. Fluency can conceal serious defects. Audit sampling should therefore examine not only whether a post-editor was assigned, but whether the records demonstrate meaningful post-editing against the approved specification.

How to Position the Service in Tenders and Client Documentation

The safest commercial approach is to describe the service precisely. State whether the project will be delivered as human translation, full post-editing of machine translation output, or another defined level of review. Avoid claiming “ISO 18587 human translation,” because that phrase combines two distinct service models and may confuse procurement teams.

Tender responses should identify the applicable standard, the role of machine translation, the level of human intervention, the competence requirements for assigned personnel, and the quality controls applied. If the client requires ISO 17100, do not assume ISO 18587 conformity is an equivalent substitute. If the client requests ISO 18587, do not respond with a general statement that all translations are reviewed by humans.

Clear contract language reduces disputes later. The purchase order, statement of work, or project specification should define intended use, language combination, content type, confidentiality requirements, terminology sources, output format, turnaround expectations, and acceptance criteria. For MTPE, it should also define the post-editing level and any restrictions on the machine translation technology used.

Audit Evidence for Each Service Model

An audit should test whether documented procedures operate in practice. Policies without project-level evidence are not sufficient. Assessors typically seek objective evidence across the service lifecycle, from enquiry and quotation through delivery, feedback, corrective action, and supplier management where external resources are used.

For human translation, useful evidence includes competence records, project specifications, translator and reviser assignments, revision records, final verification records, nonconformity handling, and client feedback. For ISO 18587, evidence should additionally demonstrate the defined MTPE workflow, post-editor competence, technical and confidentiality controls, agreed post-editing requirements, and quality evaluation methods.

Organizations that use a single generic workflow for all projects often create avoidable gaps. A controlled system should distinguish the services while maintaining common governance for confidentiality, complaints, corrective action, records retention, and continuous improvement. Internal audits should sample both human translation and MTPE projects where both are within scope.

Selecting the Right Compliance Route

The correct route depends on what your organization sells and what clients require. An LSP delivering conventional translation services should first ensure that its ISO 17100 process is fully implemented and evidenced. An LSP providing full MTPE should assess ISO 18587 requirements alongside its ISO 17100 framework, because the standards address connected but distinct controls.

Organizations should not extend their scope to ISO 18587 solely because machine translation is available in their technology stack. The standard becomes relevant when full post-editing is offered as a controlled, marketable service and the organization can demonstrate the required personnel, procedures, technical safeguards, and records.

The strongest position is not to present one model as universally superior. It is to maintain a documented decision process that selects human translation or full MTPE based on client requirements, content risk, and demonstrable capability. That approach gives procurement teams a clear answer, gives auditors objective evidence, and gives your organization a service claim it can support with confidence.