If textbook RSA is used then many attacks are possible, including dictionary attacks. This is because the value that is used in modular exponentiation (called RSAEP in the standard) is just the input decoded to a number. However, plaintext RSA is susceptible to many attacks and is not considered secure; a rather obvious one is when a message with value 0 or 1 is encrypted, as the output would be 0 or 1 as well; no attacks necessary.
Normally however a secure padding method is used. The original scheme from RSA labs was PKCS#1 v1.5 padding. This padding "wraps" the message and includes many random bits that are encrypted together with the message. These bits are not know to an attacker. They can be used to check that the decryption was successful or not. The later padding method OAEP, which is provable secure, should however be used as PKCS#1 v1.5 is vulnerable to attacks on the unpadding of the message. OAEP, just like v1.5 padding, uses random bits.
PKCS#1 reads for OAEP "Generate a random octet string seed of length hLen" where hLen is the hash function used within MGF-1, often SHA-1. That means that hLen is at least 160 bits. That means that 160 bits of randomness are encrypted together with the message. It's at least 63.95 bits for PKCS#1 v1.5 padding but usually it is more (the amount of padding depends on the difference between key and message size).
So no, you cannot use dictionary attacks because the output of the RSA encryption will always be different from the given message - if a secure padding is used. However, if for some reason the random number generator fails and leaks the not-so-random bits then you could match the padding, and perform a dictionary attack.