Insights — Market Entry Digital Infrastructure — 3 min read
Machine Translation vs Native Commercial Localisation
Machine translation has improved enormously and is genuinely fit for purpose on some pages. On the pages that carry a buying decision, it is usually the wrong tool for the job.
In short
Machine translation is generally adequate for lower-stakes, high-volume content — internal documentation, general background pages, draft content later reviewed by a human — where fluency and basic accuracy are enough. Native commercial localisation, carried out or reviewed by a native speaker with market and commercial knowledge, is needed for high-stakes commercial content — home pages, product and pricing pages, case studies and anything a buyer reads before deciding to make contact — because these pages need to make the right commercial argument for that specific market, not just read fluently.
Modern machine translation is good enough, most of the time, to be fluent and grammatically accurate — which is exactly what makes it risky to over-rely on. Fluent, grammatically correct text reads as trustworthy, which can mask the fact that it still fails to make the right commercial argument for a specific market, misses local idiom in a way only a native speaker would catch, or handles a technical or regulatory term incorrectly in a way that matters.
The honest position is that machine translation is a genuinely useful tool for some content and a genuine liability for other content, and the difference is mostly about what is at stake on the page in question, not about the general quality of the technology.
What machine translation is genuinely good at now
Machine translation tools have improved substantially and now generally produce fluent, grammatically sound text for many language pairs, particularly for straightforward factual content. For internal documentation, high-volume low-stakes content, or as a first draft that a human reviewer then refines, machine translation is a legitimate, efficient tool that would be wasteful to avoid on cost or principle alone.
Where it falls short
Machine translation translates language; it does not localise commercial argument. It will render a case study fluently without recognising that the customer name and context mean nothing to a reader in another market. It will translate a pricing page accurately without adapting currency conventions, common local payment or contracting norms, or the level of detail local buyers expect at that stage. It can mistranslate technical or regulatory terminology in ways that read fine to a non-expert but are simply wrong to a local specialist — a genuine risk in technical B2B sectors where precision matters commercially and sometimes legally.
Direct comparison
| Factor | Machine translation | Native commercial localisation |
|---|---|---|
| Fluency and grammar | Generally good for most major language pairs | Generally good, plus natural idiom and register |
| Commercial persuasiveness | Not adapted — carries over the source market's argument | Built for how the target market actually buys |
| Technical/regulatory accuracy | Risk of subtle but material errors | Reviewed by someone who understands the terminology in context |
| Cost and speed | Very low cost, near-instant | Higher cost, requires skilled human time |
| Best suited to | Low-stakes, high-volume, internal or draft content | Home page, product pages, case studies, pricing, anything pre-contact |
A sensible split, not an all-or-nothing choice
Treating this as an all-or-nothing choice wastes money in one direction or the other — either paying for full native localisation on content that does not need it, or exposing the pages that most influence a buying decision to a tool that cannot make commercial judgement calls. The practical approach identifies which content genuinely sits in front of a buying decision and applies native commercial localisation there, while allowing machine translation, ideally with a light human review pass, for everything else.
- Machine translation, unreviewed: internal documentation, low-traffic background pages, draft content for internal use
- Machine translation with human review: general informational content, lower-priority pages that still face the public
- Native commercial localisation: home page, core product and service pages, pricing, case studies, anything read before first contact
- Native localisation with technical/regulatory review: content involving specifications, standards, compliance or safety claims
AI translation tools are not the same as AI search optimisation
It is worth separating two different uses of AI here: using AI-powered translation tools to produce multilingual content, and optimising content for AI-powered answer engines. The first is a production method; the second is a distribution and discoverability question. A page can be machine-translated and still perform reasonably in AI search if its facts are stated plainly and consistently — the two issues are related but distinct, and solving one does not automatically solve the other.
Where to draw the line
The line should be drawn by commercial stakes, not by budget convenience: any page a genuinely interested buyer reads before deciding whether to make contact deserves native commercial localisation, because that is where a fluent-but-flat translation costs the most in lost conversion. Everything else can reasonably use machine translation, provided someone periodically checks it has not introduced errors that matter.
Working out what you actually need to build?
The International Digital Market Entry Report 2027 sets out the eight levels of market-entry digital infrastructure and where most companies should sensibly stop.
Related services
Written by
International Sales & Market Development Director, Evans Sales Consultancy
Published 6 September 2026 — 3 min read