As far as we can tell, Iranian.

The linguistic evidence is sparse - most of our evidence for Scythian languages consists of toponyms and personal and tribal names recorded by others. From this limited evidence, Scythian languages appear to be Eastern Iranian languages.

We know more about the languages of some of the non-Scythian peoples with Scythian cultures. Eastern Saka languages are well-known due to over 2000 surviving texts from Khotan; these were certainly Eastern Iranian languages. Sarmatian, with many known personal names, also appears to be Eastern Iranian. Alanic also appears to have been Eastern Iranian, and closely related to Sarmatian languages; modern Ossetian is often considered a descendant of Alanic (at least, it is an Eastern Iranian language and appears to be closely related to Alanic).

Thus, from the limited evidence for the languages of Scythians proper, and the better-known languages of other peoples with Scythian cultures, it's likely that all of the peoples of Scythian cultures spoke Eastern Iranian languages. Likely, but not certain. For example, the evidence for the language(s) of the Massagetae is even more sparse, with only three names that can be reliably identified as Massagetaean: the name of the people, "Massagetae" (probably Eastern Iranian), and two personal names: the queen Tomyris and her son Spargapises. These last two names are Eastern Iranian. Thus all of the evidence points to the language(s) of the Massagetae being Eastern Iranian, but no certain conclusion can be drawn from three names, especially when two of them are royal names.

On the other hand, there is no evidence at all suggesting that the Scythians, or any peoples with Scythian cultures, might have been Turkic.