Bacterial cellulosomes are generally believed to assemble at random, like those produced by Clostridium cellulolyticum. They are composed of one scaffolding protein bearing eight homologous type I cohesins that bind to any of the type I dockerins borne by the 62 cellulosomal subunits, thus generating highly heterogeneous complexes. In the present study, the heterogeneity and random assembly of the cellulosomes were evaluated with a simpler model: a miniscaffoldin containing three C. cellulolyticum cohesins and three cellulases of the same bacterium bearing the cognate dockerin (Cel5A, Cel48F, and Cel9G). Surprisingly, rather than the expected randomized integration of enzymes, the assembly of the minicellulosome generated only three distinct types of complex out of the 10 possible combinations, thus indicating preferential integration of enzymes upon binding to the scaffoldin. A hybrid scaffoldin that displays one cohesin from C. cellulolyticum and one from C. thermocellum, thus allowing sequential integration of enzymes, was exploited to further characterize this phenomenon. The initial binding of a given enzyme to the C. thermocellum cohesin was found to influence the type of enzyme that subsequently bound to the C. cellulolyticum cohesin. The preferential integration appears to be related to the length of the inter‐cohesin linker. The data indicate that the binding of a cellulosomal enzyme to a cohesin has a direct influence on the dockerin‐bearing proteins that will subsequently interact with adjacent cohesins. Thus, despite the general lack of specificity of the cohesin–dockerin interaction within a given species and type, bacterial cellulosomes are not necessarily assembled at random.