There’s little point pretending that translation hasn’t changed.
Machine translation has existed for years, but the rapid development of generative AI has made sophisticated language technology accessible to almost everyone. Businesses can now produce an English version of a French, Italian or Spanish document in seconds.
And sometimes, at first glance, it can look remarkably good.
So where does that leave the professional translator?
I think the answer lies increasingly in something that has always been central to good translation: professional judgement.
Fluent English isn’t necessarily an accurate translation
One of the most significant changes brought about by AI is that poor translation no longer necessarily looks poor.
Traditional machine translation could often be identified immediately: awkward phrasing, unnatural sentence structures and obvious grammatical errors made it clear that something wasn’t right.
AI-generated language can be much more convincing. It can be fluent, polished and entirely natural to read.
But fluency and accuracy are not the same thing.
A translation can read beautifully while subtly changing the meaning of the original, choosing the wrong interpretation of an ambiguous term, omitting a qualification or using terminology that is perfectly plausible but wrong in that particular context.
And when the English sounds convincing, those problems can be surprisingly difficult to spot.
Specialist content creates another layer of risk
This matters particularly with the kind of material I work with: financial, corporate and sustainability content.
A term in a set of financial statements isn’t interchangeable with the closest everyday English equivalent. Corporate governance has its own conventions. Sustainability reporting brings together specialist terminology from environmental reporting, employment, regulation, governance and finance.
Context matters.
So does consistency. A term may be linguistically defensible but still be inappropriate because an organisation uses an established English equivalent throughout its reporting.
Recognising those distinctions requires more than producing fluent English. It requires an understanding of both the source language and the subject matter.
The workflow is changing
For some projects, traditional human translation from source to target will remain the right approach.
For others, technology will increasingly form part of the process.
A company might use machine translation or AI to produce an initial English version and then ask a professional translator to review it. An internal team might draft content using AI and need someone to check it against the original. A translation may be largely accurate but require specialist terminology, consistency and style to be reviewed before publication.
I don’t see much value in pretending those workflows don’t exist. They already do.
The more useful question is what level of human involvement a particular piece of content requires.
This is where experience becomes valuable
Reviewing an AI-generated translation isn’t simply proofreading the English.
To review it properly, I need to read the source and translation together and ask:
- Has the meaning been conveyed accurately?
- Has anything been omitted, added or subtly altered?
- Is the terminology correct in this particular context?
- Is terminology consistent throughout the document?
- Has an ambiguous phrase been interpreted correctly?
- Does the English follow the conventions of the subject area?
- Does it actually sound natural?
Sometimes the AI output needs only minor changes.
Sometimes a sentence that initially looked perfectly good needs to be rewritten completely.
Knowing the difference is the job.
Human quality control in an AI-enabled workflow
After 18 years working professionally with French, Italian and Spanish, I don’t think my role is simply to compete with technology over who can produce English words fastest.
My value lies in being able to make informed linguistic decisions: understanding the source, recognising when something isn’t quite right, researching specialist terminology, resolving ambiguity and taking responsibility for the quality of the finished English.
That may mean translating a document from scratch.
It may mean reviewing a translation produced elsewhere.
Increasingly, it may mean acting as the experienced human specialist at the quality-control end of an AI-enabled workflow.
The tools may change. The need to know whether the final text is accurate, appropriate and ready to publish doesn’t disappear with them.
If you have an English translation of French, Italian or Spanish content – whether produced in-house, by another translator, using machine translation or with AI – I can review it against the original and bring it to publication standard.