Google translation does not work by rules and grammars. Machine translation gave up on that decades ago. Pity, because I spent well over a decade coding morphological rules for Greek, and it was a lot of fun.
Machine translation works on statistics. To gather the statistics, you need a large amount of bilingual texts.
Now, there is an order of magnitude more ancient Greek than ancient Latin texts, much of it translated. And there are a substantial number of mediaeval Greek texts as well.
But even if the interest was there in ancient Greek machine translation, the material would not be.
- Optical character recognition for the squiggles of polytonic Greek is not great, and would degrade the quality of any bilingual corpus substantially, unless someone had typed the text in. (Both Perseus and the TLG have; Google did talk to the TLG once while I was working there, but it was about teaching materials, not machine translation.)
- The classical corpus is probably not big enough to be useful for statistical machine translation; and there is a lot more bilingual text for mediaeval Latin available than for mediaeval Greek.
- Unlike Latin, the classical Greek corpus is multidialectal, which would compromise any statistics even more.
So machine translating ancient Greek would be a lot more hassle than for Latin. And because of the cultural history of Western Europe, there is much less demand for it than there would be for Latin.
Compare the number of translation requests for tattoos on this site, in Latin and in ancient Greek.