AI and the Future of Translation: Key Insights from the 2025 ATA Annual Conference

How are you using AI for translation as a business? How is it affecting your quality and workflow?

For the past few years, the topic on everyone’s minds has been that of artificial intelligence within the Translation & Interpreting industry. With so many different opinions and approaches to it, as well as misinformation and hype, I wanted to know what the experts and industry leaders had to say regarding this new technology. So, I dedicated a large portion of my time at this year's American Translators Association (ATA) Annual Conference in Boston to learning from researchers about AI, its effects on the industry, and the most efficient ways of incorporating it into a translation workflow.

The Big Picture: AI Is Reconfiguring, Not Replacing, Translators

A clear theme emerged across sessions: AI isn’t replacing translators—it’s reconfiguring how we work. Translation technology is evolving quickly, but quality still depends on human expertise, judgment, and linguistic awareness.

AI tools can assist with terminology management and speed up repetitive tasks, but they struggle deeply with stylistics and conceptual meaning. As one speaker put it, AI doesn’t do stylistics well—and that’s where translators shine.

Stylistics: The Human Touch Machines Still Miss

Stylistics is far more difficult for AI than terminology. While neural engines may correctly translate words, they often fail to grasp tone, nuance, or cultural rhythm.

For instance, a sentence like “En el informe se presentan los resultados” becomes “The report presents the results.” Accurate, yes—but it reflects U.S. stylistics creeping into Latin American Spanish, an unnatural rhythm for many readers.

That’s why professionals must focus on what makes human translators irreplaceable: stylistic intuition, conceptual understanding, and domain sensitivity.

Insights from Dr. Chela Vargas-Sierra: Context Is Everything

One of the most eye-opening sessions was led by Dr. Chela Vargas-Sierra, who broke down why AI can’t yet replace human conceptual reasoning.

“AI can’t translate concepts correctly,” she explained. “It’s basically a glossary that predicts words based on statistics. It doesn’t think for itself.” This is especially evident in legal, technical, and biomedical texts, where context determines accuracy.

Dr. Vargas-Sierra emphasized that translators must act as the AI’s guide, designing contexts that the machine can interpret effectively. She noted that AI works with embeddings—pattern recognition, not understanding—and therefore needs human input in areas like usage notes, conceptual boundaries, and register.

She outlined several critical variables for quality AI-assisted translation:

• Domain

• Communicative purpose

• Audience

• Register

• Style

For quality assurance, she recommended the MQM (Multidimensional Quality Metrics) model alongside an editor’s checklist. And above all, a human must always have the final decision.

Dr. Miguel Jimenez: The Hype vs. the Reality of AI

At his session, Dr. Miguel Jimenez of Rutgers University reminded attendees that “with most new language technologies, the hype never matches the reality.”

He described AI as “not a reflection of ourselves as language professionals, but another tool to collaborate and improve our work.” The key skill of modern translators, he argued, lies in identifying AI’s mistakes and managing its output effectively.

He also shared a striking statistic: EU-certified interpreters produce 50% fewer errors than AI systems. Humans excel at user engagement and audience reception, two areas where AI still falters. His message was clear: human-centered AI enhances, not replaces, human capacity.

James Phillips (WIPO): Strategy Before Integration

On the business side, James Phillips from the World Intellectual Property Organization (WIPO) offered a pragmatic reminder: Don’t use AI just because everyone else is.

He stressed that companies should set clear goals for AI integration, with humans embedded at every stage of the process. AI tools can boost efficiency, but only if guided by professional oversight.

His most impactful statement: “Companies that invest in quality translation make more money.” In other words, quality-driven translation, whether AI-assisted or fully human, pays off in credibility, brand trust, and international success.

Takeaways: The Future Is Human-Centered AI

Leaving the ATA Conference, one message resonated loudest: AI is transforming workflows, not eliminating the need for human translators and interpreters.

Translators are evolving into language designers and context architects, ensuring meaning, tone, and purpose survive the translation process. AI can automate tasks, but not understanding. It can predict words, but not meaning.

So, the real question isn’t “Will AI replace translators?” but rather “How can translators lead the evolution of AI in language?”

Key Takeaways for Translation Businesses in 2025

• Use AI strategically, not reactively.

• Always embed a human reviewer in your workflow.

• Focus on stylistics and conceptual quality—areas where humans outperform AI.

• Apply the MQM and editor’s checklists for quality assurance.

• Train your translators to become AI supervisors, not AI users.

Author: Oliver Flynn