Does companion planting actually work, or is it folklore?

Some of it is well-supported and some of it is not, and the two get repeated together with equal confidence. It is worth separating them.

The mechanisms with reasonable evidence behind them are physical and ecological rather than mystical. Tall plants shading heat-sensitive ones works — lettuce sown in the shadow of climbing beans genuinely bolts later. Flowering plants such as calendula, phacelia and umbellifers do measurably increase populations of hoverflies and parasitic wasps, which reduces aphid pressure. Ground-covering crops such as squash suppress weeds around slower-establishing plants. And legumes do fix nitrogen, though most of it becomes available to the following crop when the roots break down rather than to the neighbour this season.

Trap cropping is also real: nasturtiums do attract blackfly away from beans, provided you are willing to sacrifice the nasturtiums.

The weaker claims are the specific pairings passed down as rules — that basil planted beside tomatoes improves their flavour, or that particular species must never be adjacent. Controlled trials have generally not supported flavour effects, and most "enemy" pairings turn out to be simple competition for light, water or root space rather than any chemical antagonism.

The genuinely useful principle underneath all of it is diversity. A bed with several unrelated crops interplanted is harder for a specialist pest to find and exploit than a solid block of one thing, and it uses light and root depth at different levels. That effect is real and does not depend on any specific pairing.

So plant a mixture, include flowers, avoid putting two heavy feeders of the same family in the same spot two years running, and treat the more specific folklore as harmless rather than essential.