The Actual Process
Start with the Lewis structure. Draw it properly, including all lone pairs. Most people skip lone pairs and wonder why their answer is wrong six months later when they actually need it for something that matters. Once you have the structure, identify the geometry around each central atom using VSEPR theory. This gives you the shape. Then look at the electronegativity differences between atoms. A bond is polar if the electronegativity difference is greater than about 0.4 on the Pauling scale. Individual bonds can be polar without the whole molecule being polar. The question everyone actually needs to answer is whether those bond dipoles cancel out. Draw the dipole vectors along each bond, pointing toward the more electronegative atom. If the vectors sum to zero, the molecule is nonpolar. If they don't, it's polar. This vector addition step is where people lose points on exams and mess up in the lab.
How To See If A Molecule Is Polar Or Nonpolar
I spent an entire semester teaching gen chem and the same mistakes came up every single term. The most common one is assuming that if a molecule contains polar bonds, it must be polar overall. CO2 is the classic trap. Both C=O bonds are quite polar, but the linear geometry makes the dipoles point in exactly opposite directions. They cancel. Net dipole is zero. The molecule is nonpolar despite having two strongly polar bonds. Another thing nobody gets right away: symmetry alone isn't enough. You have to consider whether the symmetric arrangement actually cancels the dipole moments. CCl4 is tetrahedral and nonpolar because all four bonds are identical. But CH3Cl is also tetrahedral and it's polar because one of the substituents is different. The symmetry is broken. Here's a case that caught me off guard once. A student brought me SF4 and was convinced it was nonpolar because sulfur is in the middle and everything looked "balanced" on paper. SF4 has a seesaw geometry due to one lone pair on the sulfur. The lone pair sits in the equatorial position, which means the axial fluorines don't perfectly cancel the equatorial ones. The molecule has a nonzero dipole moment of about 0.632 D. I had to walk through the actual 3D arrangement three times before they saw it. If you're not visualizing the geometry in three dimensions, you're guessing.
Ozone is another one that trips people up repeatedly. O3 has a bent shape with a resonance structure that gives it a formal charge separation. It's polar. The central oxygen is sp2 hybridized and the molecule has a dipole moment of roughly 0.53 D. The dipole doesn't point along any single bond. It points somewhere between the two bonds, toward the terminal oxygens, modified by the lone pair on the central atom.
Get the Full Details

Edge Cases and Where the Simple Method Breaks
The vector cancellation approach works fine for small molecules where you can draw the geometry by hand. It starts falling apart with larger organic molecules that have multiple chiral centers or flexible conformations. Cyclohexane derivatives are a good example. In their chair conformations, substituents can be axial or equatorial, and the dipole moment changes depending on which conformer dominates at room temperature. You can't just look at a flat 2D drawing and call it. Metal complexes are even worse for this kind of analysis. Take something like [Co(NH3)4Cl2]+. Depending on whether it's the cis or trans isomer, one is polar and the other isn't. Drawing the Lewis structure doesn't tell you which isomer you have. You need to know the actual coordination geometry and the spatial arrangement of ligands. Crystal field theory and ligand field theory are what you actually use in practice, not dipole vector addition. Solvent effects are another area where the simple textbook model fails completely. A molecule that appears nonpolar in a vacuum might behave differently in solution. Acetone has a significant dipole moment in the gas phase, but when you dissolve it in water, the surrounding solvent molecules reorient and partially screen the dipole. The measured dipole moment in solution can differ from the gas-phase value by a substantial amount depending on the solvent polarity and concentration.
There's also the issue of dynamic molecules. BF3 is planar and nonpolar. But if you substitute one fluorine with chlorine to make BFC l2, the molecule is still trigonal planar, yet now it's polar because the three bonds are no longer equivalent. The dipole moment is about 1.07 D. Students often forget that replacing one atom in a symmetric molecule destroys the symmetry immediately.
Practical Shortcuts That Actually Work
If you want to quickly determine polarity without drawing every vector, memorize the geometries that guarantee nonpolarity when all surrounding atoms are identical: linear (AX2), trigonal planar (AX3), tetrahedral (AX4), trigonal bipyramidal (AX5), and octahedral (AX6). Any deviation from these — a lone pair, a different substituent, or a distorted geometry — usually means the molecule is polar. The exceptions are rare enough that this heuristic covers probably 90 percent of cases you'll encounter in an undergraduate course. For larger molecules, look for a center of inversion. If a molecule has one, it cannot be polar. This is a quick group theory shortcut that saves time on things like benzene, ethylene, and square planar complexes. SF6 has a center of inversion and is nonpolar. trans-[PtCl2(NH3)2] has one too and is nonpolar. The cis isomer does not and is polar. When you need actual numbers instead of yes-or-no answers, you look up the dipole moment in debyes. Water is 1.85 D. Ammonia is 1.47 D. HCl is 1.08 D. These values are measured experimentally and you shouldn't rely on estimates when precision matters. A dipole moment above about 0.5 D is generally considered polar for most practical purposes in organic chemistry and biochemistry.

The hardest molecules to classify are things like 1,2-dichloroethane where free rotation around the C-C bond creates multiple conformers. The anti conformer is nonpolar because the two C-Cl dipoles cancel. The gauche conformer is polar. At room temperature, both exist in equilibrium and the measured dipole moment is a weighted average. Saying the molecule is simply "polar" or "nonpolar" without specifying conditions is technically incorrect, though most introductory courses won't penalize you for calling it polar.