Theorem. Let anxn+⋯+a1x+a0=0 have integer coefficients, an=0, a0=0. If p/q (lowest terms, gcd(p,q)=1) is a root, then p∣a0 and q∣an.
This turns an unguided, infinite guessing problem into a short, finite list of candidates: list every divisor of the constant term a0 as a possible numerator, every divisor of the leading coefficient an as a possible denominator, form every resulting fraction, and test only those by direct substitution.
Special case (monic equations). If an=1, the theorem forces any rational root's denominator to be ±1 — so the root must be an integer, and specifically a divisor of a0. A monic integer-coefficient polynomial can therefore never have a genuinely fractional rational root.
The theorem only narrows the search — it does not guarantee any candidate actually works (e.g. x2+4=0's candidates ±1,±2,±4 include no real root at all, since both roots are imaginary), nor does testing every candidate necessarily find all the roots if some are irrational or non-real.
Worked illustration. For 2x3+3x2+2x+3=0: an=2,a0=3 give candidate numerators ±1,±3 and denominators ±1,±2, so the full candidate list is ±1,±21,±3,±23. Testing shows −23 is the only rational root; dividing (2x+3) out leaves x2+1=0, giving the remaining roots i,−i.