Mercenary participants — users who engage solely to qualify for an airdrop and exit immediately — are impossible to eliminate entirely. But you can reduce their share of the allocation by weighting criteria that are expensive to fake: long-duration activity (months, not days), genuine product usage (not just token transfers), and social / identity signals that link on-chain behavior to real humans.

Retroactive airdrops with unannounced snapshots are more effective than prospective ones with public criteria, because they reward behavior that occurred before any incentive was known. The tradeoff is that you cannot use the airdrop to drive new behavior. Most projects benefit from a combination: a retroactive component for early adopters and a prospective component for targeted growth.