Structure.

An AI assistant wraps code in extra functions, classes, or interfaces that add a layer of indirection without hiding any complexity or enabling reuse — abstractions that merely forward to a single call site.

AI assistants tend to paste a fresh, slightly-tweaked copy of existing logic instead of reusing or extracting a shared function, inflating duplicated code while refactoring quietly disappears.

AI assistants pile new, near-duplicate code next to what already exists instead of restructuring it, so duplication climbs and refactoring — the rework that keeps a codebase healthy — quietly disappears.

The model hand-writes bespoke code for something a standard library, an existing dependency, or a helper already in the repo already does — adding duplicated, less-tested logic instead of calling what is already there.