"Our quality assurance team found that most API-delivered translations originate as Google Translate translations that are slightly edited to make some sense, then delivered to clients who have no idea that they just received a bad translation."
Using free translation tools means using automated, machine translations. While machine translation has improved over the years, it's still a toddler. It makes strides that look like upright walking, but it falls over and over. Most translation tools offered online are good at translating greetings, and very simple sentences that are semantically obvious. Where they fail is at the level of pragmatics and sociolinguistics. This means that a computer processing zeros and ones has NO clue what the human-emotional-situational-culturally-expectant context is unless the algorithm has uncovered specific patterns that it has been programmed to pick up. As much as people like to believe that language processing algorithms are "living" in some way, they aren't. Machine learning researchers openly admit this failing, and there is currently no way to plan for the trillions of situations possible, many of which are physical or irrational, and not written on a page. Thus, the notion of "garbage in, garbage out" in language learning often changes to "good in, garbage out" with free translation tools. But the work of translation tool developers is very important, and the garbage out can be very funny, as you see in the videos below. In short, translation tools should never be relied on for serious communication.
The least bad automatic, free translation tools are as follows: