This conversation analytic study debates how participants in recorded mediator-(prospective) client encounters, orient to their own or recipients’ talk as potentially biased, now and then plainly marking that talk as racist (sexist, etc.). Mediator responses show two wide categories, either removing (e.g. via reformulation) or objecting the -ism (e.g. via warning). Both imply alignment error or disaffiliation instead of the mediation-instructed neutral stance.