How do you make a chess bot play badly in a believable way?

A developer is seeking advice on creating a chess bot that plays poorly in a believable manner, rather than feeling like it's deliberately losing. The current setup, which combines Maia and Stockfish with custom move-selection rules, results in an inconsistent opponent that sometimes feels like two different players. The developer is looking for ways to give AI believable blind spots and continuity of thought, drawing on experiences from chess or other games.

This post uniquely highlights the challenge of making AI perform imperfectly yet believably, unlike typical AI development focused on optimal performance.