A buyer asks whether a project will receive human revision, while the delivery team plans machine translation post-editing. That apparent equivalence can create a material compliance issue. ISO 18587 versus human revision is not simply a choice between two ways of checking translated text. It is a distinction between defined service processes, assigned competencies, documented requirements, and the evidence an organization can present during an audit or tender assessment.

For language service providers, the correct answer begins with the client specification and the service scope. A human review step may be appropriate for conventional translation workflows. It does not automatically demonstrate conformity with ISO 18587, which addresses the full post-editing of machine translation output. Conversely, an ISO 18587 workflow should not be described as generic “revision” if the terminology used in a contract, statement of work, or quality procedure has a more specific meaning.

What ISO 18587 Covers

ISO 18587 specifies requirements for the full post-editing of machine translation output. Its purpose is to establish a controlled process through which machine-translated content is edited so that the final output can be considered comparable in quality to human translation.

The standard does not certify a machine translation engine, a translation management system, or a language pair. It addresses the service provider’s ability to manage a post-editing service. This includes agreement of project specifications, assessment of source content suitability, use of appropriate technical resources, assignment of competent post-editors, and quality management of the delivered target content.

A central point is that ISO 18587 concerns full post-editing, not light post-editing. Light post-editing may be commercially valid when the intended use permits understandable but non-polished output. It is not the service level addressed by ISO 18587. Where a client requires publication-ready, legally sensitive, regulated, customer-facing, or contractual content, the difference must be explicit before production starts.

The post-editor works from machine translation output and corrects it to meet the agreed specifications. The process must address accuracy, completeness, terminology, linguistic conventions, formatting, and the intended purpose of the target text. The organization must also control the conditions under which machine translation is used, including confidentiality, data handling, and technical suitability.

What Human Revision Means in a Translation Workflow

Human revision normally refers to a bilingual examination of a translation by a person other than the translator. In an ISO 17100-aligned workflow, revision is performed against the source text and is a defined quality assurance activity. The reviser checks whether the translation is fit for purpose and identifies errors in meaning, terminology, language, and other agreed requirements.

That process starts from a human-produced translation, although technology may support the work. The distinction is operationally significant: the original translator creates the target text, and an independent reviser assesses it. In a machine translation post-editing workflow, the post-editor develops the final text from machine-generated output. The initial production method, role design, effort profile, and associated risk controls are therefore different.

Human revision can also be used more broadly in commercial language services. Some clients use the term for monolingual proofreading, stylistic editing, desktop publishing checks, or a final review by a subject-matter specialist. These activities may add value, but they should not be represented as bilingual revision unless the reviewer actually compares source and target content.

ISO 18587 Versus Human Revision: The Core Difference

The practical difference is not whether a qualified human is involved. Both processes require qualified personnel. The difference is what the human professional is asked to do, what input they receive, and what process the provider must control.

Under ISO 18587, the relevant input is machine translation output. The post-editor must bring that output to the quality level defined in the project specification. The provider needs procedures covering the machine translation and post-editing process, including evaluation of project suitability and appropriate handling of source and target data.

In human revision, the reviewer examines a translation that has already been produced by a translator. The independence of the revision role is a key control where ISO 17100 requirements apply. A post-editing process may include additional review steps, but it should not be assumed that post-editing itself is identical to independent revision.

This distinction becomes especially important in procurement documentation. If a tender requires ISO 18587 conformity, a supplier that offers only conventional translation plus revision may not satisfy the requirement. If a buyer requests “human revision” following machine translation, the supplier must clarify whether the buyer expects full post-editing, independent bilingual revision after post-editing, or both. The quotation, workflow description, and quality plan should reflect the agreed answer.

Competence and Role Assignment

A compliant process depends on more than a quality claim. ISO 18587 requires post-editors to have relevant translation competence and post-editing competence. They must be able to recognize the characteristic errors and limitations of machine translation, apply the client’s specifications consistently, and use the designated technology effectively.

For management teams, this requires a documented competence framework. Personnel records should demonstrate qualifications, professional experience, language and subject-matter capability, and evidence of post-editing training or assessed competence. A linguist who is qualified to revise human translation is not automatically qualified for every post-editing assignment, particularly in technical, medical, legal, financial, or highly regulated content.

Role assignment should also consider the risks created by the machine translation environment. Engine behavior can vary by domain, language direction, content type, terminology coverage, and source-text quality. The provider should avoid assigning work solely on the basis of language availability or throughput targets. A documented qualification and allocation process is more defensible in both external audits and client due diligence.

Documentation That Auditors and Buyers Expect

Organizations often have capable linguists but insufficient evidence that their process is controlled. Certification assessments examine whether documented procedures are implemented consistently, not merely whether a sample translation appears acceptable.

For ISO 18587 services, audit evidence commonly includes the client agreement or project specification, confirmation of service level, suitability assessment records where applicable, competence records for post-editors, project instructions, terminology resources, quality control records, and records of nonconformities or corrective action. The organization should be able to show how confidentiality and data protection are managed when machine translation technology is used.

Where human revision is part of the workflow, the project record should identify the translator and reviser, establish the relevant competence of each role, and show that revision was performed as required. Review comments, tracked changes, quality reports, or workflow status records may support this evidence, provided they are retained under a controlled records procedure.

The objective is traceability. An auditor should be able to follow a project from the accepted client requirements through production, quality control, delivery, and any post-delivery action. A generic statement such as “all work is reviewed by humans” does not provide that traceability.

Choosing the Appropriate Service Model

The service model should be selected according to risk, intended use, client requirements, and commercial constraints. Full post-editing under ISO 18587 may be suitable where machine translation can be used responsibly and the client requires target text quality comparable to human translation. Conventional translation followed by revision may be more appropriate for content with high creative, legal, safety, reputational, or confidentiality sensitivity, or where machine translation is prohibited by contract.

There are also cases where an additional independent review after full post-editing is justified. This may apply to regulated submissions, critical product information, or projects with unusually complex source material. That additional control should be defined as a separate workflow step, with a clear purpose and assigned responsibility. It should not be assumed to be included merely because the service is described as ISO 18587-compliant.

Cost and turnaround time remain relevant, but they should not drive the classification of the service. Mislabeling a light post-editing project as full post-editing, or presenting post-editing as conventional human revision, can expose the provider to contractual disputes and nonconformities during assessment.

Building an Audit-Ready Process

A practical implementation begins by separating service descriptions in the quality management system. Define conventional translation and revision, full machine translation post-editing, any light post-editing offering, and any additional review activities. Align sales terminology, quotations, project instructions, supplier requirements, and production workflows with those definitions.

Next, test whether project managers can apply the procedures consistently. They need decision criteria for machine translation suitability, escalation routes for problematic source content, and a method for recording client approvals or restrictions. Quality managers should review whether post-editing records demonstrate the same discipline as other controlled language-service processes.

Internal audits are particularly useful when they examine real project files rather than only procedures. A file review can reveal whether the agreed service level was correctly identified, whether the assigned personnel met competence requirements, and whether the final quality control was proportionate to the project risk. Findings should lead to corrective action, not informal workarounds.

The most reliable position in a tender or certification audit is not to claim that one process is universally better. It is to demonstrate that the organization can distinguish ISO 18587 post-editing from human revision, select the appropriate service deliberately, and provide objective evidence that the promised process was delivered.