One of my friends who does Spanish tutoring also does “clean up” of automated translations to make them sound native-speaker like. Según her, DeepL is not bad but not great. If you use an automated translation for important communications, native speakers will be able to tell and may even be offended that the company did not bother to use the services of a quality translator.
One example from the NYTimes.
English: https://www.nytimes.com/2018/09/28/style/modern-love-how-i-m...
Spanish: https://www.nytimes.com/es/2018/09/28/modern-love-donador-es...
Excerpt #1
Original English:
> A couple of years ago I began seeing ads for 23andMe, a service that analyzes your saliva — you spit in a test tube and mail it off for analysis — and provides you with information about ancestry, health and DNA relatives.
Translated Spanish via DeepL:
> Hace un par de años comencé a ver anuncios de 23andMe, un servicio que analiza tu saliva - escupes en un tubo de ensayo y lo envías por correo para analizarlo - y te proporciona información sobre ascendencia, salud y parientes ADN.
Translated Spanish (professional translator):
> Hace un par de años, comencé a ver anuncios de 23andMe, un servicio que analiza tu saliva —la pones en un tubo de ensayo y lo mandas por correo para que se analice— y te da la información de tus ancestros, salud y familiares según tu ADN.
DeepL’s translation has the same faults as other services. Namely, in Spanish, object pronouns are very important. In the sentence describing how he sent a saliva sample, the DeepL translation does not use pronouns correctly like a native speaker would. Also, it stumbles on what is the most common vernacular. The DeepL translation does not use the right way to say “DNA relatives” which is “familiares según tu ADN”. “Según” is commonly used to denote authority of a source or in other words “according to”.