We introduce a generating-function framework for analyzing the no-feedback card-guessing game after $k$ Gilbert–Shannon–Reeds riffle shuffles. We show that the distribution of the card appearing in position $i$ can be expressed as a structured mixture of $2^k$ tractable components, each corresponding to a sum of independent Bernoulli trials. From this decomposition, we derive an explicit closed-form expression for the probability generating function, represented as a product of binomial-type polynomials with a clear and systematic structure, valid for any number of cards $n$ and any number of shuffles $k$. This formulation replaces recursive convolutions with a single analytic expression, enabling efficient computation and revealing the combinatorial–probabilistic structure underlying riffle shuffles. Beyond exact evaluation, the framework connects optimal no-feedback strategies with the generating functions and suggests asymptotic behavior in both the fixed- $k$, large- $n$ and fixed- $n$, large- $k$ regimes.