Is KMAC just SHA-3-256(KEY || message)

Standard KMAC is more than that thanks to domain separation prefixes; NIST SP 800-185

KMAC128(K, X, L, S):

Validity Conditions: $\text{len}(K) < 2^{2040}$ and $0 \le L < 2^{2040}$ and $\text{len}(S) < 2^{2040}$

1. newX = bytepad(encode_string(K), 168) || X || right_encode(L).
2. return cSHAKE128(newX, L, “KMAC”, S).

and

cSHAKE128(X, L, N, S):

Validity Conditions: $\text{len}(N) < 2^{2040}$ and $\text{len}(S) < 2^{2040}$

1. If N = "" and S = "": return SHAKE128(X, L);
2. Else: return KECCAK[256](bytepad(encode_string(N) || encode_string(S), 168) || X || 00, L)

and NIST FIPS 202

SHAKE128(M, d) = KECCAK[256] (M || 1111, d),

NIST's information about suffixes;

The suffix supports domain separation; i.e., it distinguishes the inputs to KECCAK[c] arising from the SHA-3 hash functions from the inputs arising from the SHA-3 XOFs defined in Sec. 6.2, as well as other domains that may be defined in the future.

I.e words, we have different random oracles per domain separation. SHA3 and Blake2 are more close to random oracles than SHA256 since they are free from the length-extension attack.