Suppose you tag a post with eagle. Eagles are simply a type of bird, so ideally, you would like people who search for bird to see your eagle post. You could tag your post with both eagle and bird, but this starts to get tedious after a while.

Tag implications can be used to describe is-a relationships. An eagle is a type of bird. When an eagle → bird implication is created, then whenever someone tags a post with eagle, e621 will also tag it with bird.

Tag implications have a predicate and a consequent. The predicate is what is matched against. In the previous example, it would be eagle. The consequent is the tag that is added. In the example, it would be bird.

You can have multiple implications for the same predicate. e621 will just add all the matching consequent tags. For example, if we created a eagle → avian implication, then anytime someone tagged a post with eagle it would be expanded to eagle bird avian.

Implications can also be chained together. Instead of eagle → avian we could create a bird → avian implication. The end result would be the same.

This implication process occurs AFTER the alias process.

It's easy to go overboard with implications. It's important not to create implications for frivolous things; for example, we could theoretically implicate everything to an object tag, but this is pointless and only adds bloat to the database. For cases where the predicate and the consequent are synonymous, aliases are a much better idea as they have lower overhead.

While you can suggest new implications, only an administrator can approve them.