Is machine translation good enough to use as is?

For gist, often yes. For anything published or high-stakes, no: it should be post-edited by a qualified human, because it cannot detect its own errors.

What is the difference between machine translation and AI translation?

They overlap. Machine translation is the broader term for automatic translation; AI translation usually emphasises that modern engines are neural or built on large language models. In practice they refer to the same thing.

AI and machine learning

Machine translation

Machine translation is the automatic translation of text by software, without a human translating each sentence. Modern systems are built on neural networks and are fluent enough that, for many kinds of content, editing their output is faster than translating from scratch.

How it works

A machine translation engine takes source text and generates a translation without a human translating each sentence. Modern engines are neural: trained on large volumes of human-translated text, they learn statistical patterns and produce output one token at a time.

They are fast, cheap and available on demand, and for many content types their output is a good enough first draft that editing it beats translating from scratch. What they cannot do is know when they are wrong, which is why output for publication is post-edited.

How SourceTarget uses it

SourceTarget uses machine translation as the drafting engine in its post-editing workflow, choosing the engine that performs best for the content and language direction. Every job carries confidence and quality scoring so the raw output is never mistaken for a finished translation.