Are ISO 17100 and ISO 18587 Still Relevant in Today’s AI-Driven Translation Landscape?

In the translation and localization industry, ISO 17100:2015 and ISO 18587:2017 are two of the most recognized and widely adopted standards. ISO 17100 sets the requirements for the basic translation process—project management, translator qualifications, revision, and delivery—to ensure consistency and professionalism in language services. ISO 18587, in turn, defines the requirements for machine translation post-editing and provides guidance on how human linguists should edit, revise, and refine machine translation output.

Despite their widespread use and high certification rates among language service providers, these standards have faced criticism, with some considering them outdated. In fact, ISO 18587 is currently under revision, with the new version expected in 2026. Additionally, two non-certifiable guidance standards were published in 2024 that aim to complement or compensate for the shortcomings of the earlier standards. Firstly, ISO 5060, which provides a framework for objectively and measurably assessing the quality of human or machine-generated translations. Secondly, ISO 11669, which focuses on developing clear specifications for translation projects.

Are These Standards Still Relevant?

The obvious question, then, is whether ISO 17100 and ISO 18587 remain relevant. Do they provide value to clients and to the organizations that certify in them? What benefits do they offer? Are their limitations insurmountable?

In this article, we attempt to analyze these questions based on our own experience as a language service provider that has been certified in both standards for three consecutive years. We hope to share the lessons learned throughout the process—from certification to surveillance audits—from a practical perspective of production efficiency within today’s AI-enabled workflows.

Criticisms from the Field

Some localization professionals and practitioners argue that ISO 17100 is obsolete because it was drafted before generative tools and large-scale machine translation emerged. They see its linear, process-oriented design as too rigid for agile or tech-driven teams.

Many also claim that ISO 17100 certification does not necessarily guarantee quality, since the standard prioritizes documentation and procedures over measurable results. Moreover, it lacks mechanisms to assess translation performance or client feedback.

Others believe that ISO 18587:2017 is likewise outdated. It was written with early neural machine translation systems in mind, without considering the new generation of large language models or iterative AI tools. Critics argue that ISO 18587:2017 offers a narrow view of the post-editor’s role, focusing mainly on error correction rather than on the broader and increasingly necessary human-machine collaboration, and that it provides no objective quality evaluation metrics.

The Subjectivity Challenge

ISO 5060 appears to address the long-standing issue of subjectivity in translation quality evaluations (“I just don’t like this version”), which is based on a holistic, impressionistic approach without precise metrics. This standard provides a framework for evaluating the quality of human translations, machine translation post-editing output, and even raw machine translation through more objective methods and tools such as scorecards, error classification, severity levels, and evaluator competence. The new ISO 11669, on the other hand, helps clients define very clear translation specifications to reduce ambiguity.

It is important to clarify that these standards are informative (non-certifiable). This means that organizations often claim to follow, align with, or implement ISO 5060/11669 guidance—or advertise “ISO 5060-level QA”— rather than holding an ISO-issued certificate.

Since many clients are unfamiliar with these standards, such claims may serve marketing purposes more than reflect reality. There is also a risk of “ISO-washing”: claiming compliance without applying the standard in practice.

Our Certification Journey

Obtaining a certification in ISO 17100 and ISO 18587 requires work before, during, and after the certification process.

In our case, we completed a six-month preliminary consultancy during which we reviewed each point of the standards and reflected on how we carried out each process. This exercise is extremely valuable for any organization. As human beings, we tend to resist change and believe we are already doing everything well. However, by deeply examining and questioning your operations, you realize there is always room for improvement—which we accomplished using existing resources, time and staff.

In our experience, the improvements extended beyond processes and efficiency to quality overall (with fail rates under 1%). We complemented this with a kaizen (continuous improvement) approach which we adopted across the organization: employees themselves propose improvements within their specific areas.

Audits demonstrate compliance and commitment to clients, serving as a guarantee of the organization’s integrity. Auditors are external third parties whose task is to detect nonconformities, deviations, or other issues related to the audited standards. They ask to look at projects, examine documented evidence, and may interview team members. This means that the entire organization must work consistently and that its practices must match internal written processes. Surveillance audits ensure regular oversight and require the organization to stay up to date as changes and improvements occur.

Assessing Relevance

ISO 17100 and ISO 18587 are somewhat rigidity. In our view, they do not cover all current services or workflows: they are oriented toward direct-client services while in Latin America most language service providers also work for multilingual language vendors (MLVs). They may also differ from real-world practices in other ways. Ultimately, they can be improved. Considering that more than three years may pass between the proposal and the publication of an ISO standard, they will likely always lag behind ongoing developments, especially during periods of rapid change. However, it is important to remember that they address core aspects, draw on existing best practices, and are developed by committees of experts from various sectors.

Organizations have the flexibility to adapt their processes to their own needs and workflows. In fact, they can expand upon these standards to reflect their real practices. It would be naive to expect a single standard to satisfy everyone. Auditors check not only for compliance but also for process effectiveness. Auditors and auditees must rely on judgment and common sense to ensure that documented processes reflect real work and help optimize it. An auditor may even suggest improvements that are not directly related to the text of the standard.

Quality in High-Volume AI Workflows

It is often argued that these standards do not necessarily guarantee quality or provide objective metrics for evaluation. However, such evaluations are necessary in only a very small portion of mainstream language services. In most large-scale, AI-automated production workflows, speed, volume, and price take precedence over quality (“fit-for-purpose” quality). Human quality evaluations are impractical in these environments due to their slowness and cost, whereas QA tools like Xbench provide a faster and more cost-effective alternative.

Moreover, ISO 5060 is inspired by practices and methodologies that are already used in the market for accounts with high quality expectations. In our case, as early as 2019—before obtaining our certification—we began working on various marketing and IT accounts subject to Language Quality Evaluation (LQE) processes. In these processes, an external evaluator marks errors of different severity levels on scorecards and a minimum passing score must be achieved. While this methodology dispels the idea that “revisers are subjective,” it is a highly manual process that contrasts sharply with today’s automated landscape.

Final Thoughts

Although ISO 17100 and ISO 18587 do not fully capture the complexity of today’s AI-driven workflows, they are valuable frameworks for structuring processes, professionalizing teams, and demonstrating a commitment to verifiable best practices. Their greatest contribution lies in providing a solid and adaptable foundation that each organization can expand according to its reality. When combined with more recent standards like ISO 5060 and ISO 11669, they remain relevant as part of a continuously evolving quality ecosystem.