Quick answer
A biased survey question nudges people toward an answer or makes honest answers hard to give. The most common types are leading, loaded, double-barreled, absolute and vague questions, plus unbalanced or incomplete answer options. Fix them by removing hinting adjectives, asking one specific thing at a time, screening before follow-ups, balancing scales and randomizing option order.
What makes a survey question biased?
Survey answers are supposed to describe the people who give them. A question is biased when the answers start to describe the question instead: its wording, its assumptions, its answer options or its position in the survey. Researchers call the result response bias, and it is different from sampling error. A bigger sample does not fix it; a better question does.
Bias usually creeps in by accident. The author knows what the product is meant to do, hopes the new feature is liked, and writes a question that sounds natural to them. Respondents, who want to be helpful and finish quickly, follow the cues. The examples below show the patterns to look for. For the positive rules behind them, see how to write survey questions.

