Algorithmic bias is a systematic skew in a software system's outputs that can create unfair or misleading outcomes. It can arise from unrepresentative data, flawed labels, design choices, or the conditions in which a system is used. In AI systems, retrieval and evaluation choices can contribute as well as model training.
Bias is not the same as a single factual mistake or ordinary variation between answers. Establishing a pattern requires a relevant comparison and enough evidence to separate a consistent skew from chance.
Practical relevance
An assistant tested only with English-language questions about large US vendors may give an incomplete picture of the options available to buyers in another market. The measurement itself can also be biased if its prompt set excludes those buyers' needs.
Compare relevant markets, languages, and customer situations using a balanced set of questions. A missing brand alone does not prove bias; source coverage, product fit, and retrieval failures are alternative explanations worth checking.