Although fdb is correct that a symbol like *bh should not be taken to imply any specific phonetic content, these phonemes obviously had such content, and it's possible to speculate about what it was. The facts that lead people to think that the PIE "voiced aspirates" were breathy voiced stops as in Indic have to do with their reflexes in the daughter languages, namely:
- In Indic, they appear as breathy voiced stops. (In some modern Indic languages they remain such, in others they have become plain voiced stops, and in some Gypsy dialects they have become voiceless aspirates, as in ancient Greek.)
- In Greek, they appear as voiceless aspirated stops.
- In several branches (Anatolian, Celtic, Balto-Slavic, Iranian, Germanic, Armenian), they appear as voiced stops.
- In Italic, they appear to have first changed to voiceless fricatives, with various later developments.
This suggests that:
- They were stops. (A change of stops to fricatives in Italic is much more likely than a change of fricatives to stops in all the other languages.)
- They were voiced. (They're voiced in the large majority of daughter languages; also, Indic has a separate voiceless aspirate series with which they contrast.)
- They seem, on the evidence of Indic and Greek, to have had some kind of phonation feature that distinguished them from the plain voiced stops. They must have been distinct from these in some way (since they show different reflexes in about half the branches), and the distinction seems to have had to do with a more open glottis state.
Putting all these features together, you arrive at breathy voiced stops. That said, this conclusion is far from unproblematic: first, because only one sub-branch (Indic) then actually preserves the original sounds into the historical period, which is maybe a bit strange; second and more importantly, because the system of contrasts that you end up with for PIE stops (voiceless / voiced / breathy voiced) is cross-linguistically either extremely rare or nonexistent. So it's still an unsolved problem.