Suppose Alice wants to convince Bob that ciphertext c = (a, b) = (gr, m*hr) is some properly encrypted plaintext (not just random numbers). Obviously she can use zk-proof of discrete logarithm knowledge for the first part of the ciphertext (a = gr). But what can she do with b-part?
One of my ideas: if m belongs to subgroup G (with a generator g) than we should prove that b belongs to G. But it's just a new problem.
Alice can also use exponential ElGamal (c = (gr, gm * hr)) if it makes things easier.