While research on conspiracy theories has rapidly expanded in recent years, it remains methodologically narrow in psychology. Most studies rely on self-administered questionnaires that treat survey responses as stable indicators of belief (“x% believe in conspiracy y”), implicitly assuming that such beliefs are stable individual dispositions, detached from social and historical context. This paper challenges that assumption by approaching conspiracy beliefs within context-dependent communicative practices. Drawing on a mixed-methods study of responses to COVID-19, in which participants first completed a questionnaire measuring endorsement of specific conspiracies, followed by interviews inviting them to elaborate on their earlier responses. The comparison between these two contexts of data collect reveals striking divergences that illuminating constraints in the method. Factors shaping these divergences include (1) the shift from written to oral communication; (2) social desirability and the self-reflexive distancing of statements such as “I know this is a conspiracy theory, but…”; (3) the openness of trust-based conversation; (4) differing interpretations of Likert items as hypothetical possibilities versus personal convictions; and (5) the contrast between forced categorical choice and an expression of a range of possible belief in dialogue. Taken together, the findings problematize the conception of “the conspiracy theorist” as a single, monological type. Instead, they invite an understanding of conspiracy belief as relational, being continuously negotiated within specific communicative and societal moments.