The central question that guides this paper is whether generative AI should be included in the Digital Markets Act (DMA) as a separate core platform service. Our preliminary answer is in the affirmative. Certain generative AI applications might already be indirectly caught by the DMA when current gatekeepers integrate proprietary or third-party generative AI into their core platform services. Their standalone generative AI offering might also be caught by some of the DMA obligations under Articles 5 and 6. However, without the explicit inclusion of generative AI as a form of core platform service, any protection provided by the DMA can only have limited effect. There are growing concerns of gatekeeping in the generative AI industry, and such indirect application does not address the contestability and fairness issues specific to generative AI.This paper is structured in four parts.(i) Platformisation of generative AI:Generative AI models are increasingly acting as gateways between businesses and users, as defined by the DMA. More specifically, certain types of generative AI models emerge as novel types of digital platforms, fundamentally different from existing platforms, in two primary ways: generative AI as platform infrastructure, and generative AI that augments existing services.(ii) Gatekeeping characteristics: We find that certain companies that develop large generative AI models strongly exhibit four characteristics which facilitate the emergence of gatekeepers: (i) first-mover advantage, (ii) capital-intensive resources, (iii) data resources, and (iv) system integration. First, we illustrate how a first-mover advantage functions in the generative AI sector through setting industry benchmarks and rapid scaling strategies. The operative distinction here is that a first developer of a generative AI model may not necessarily benefit from a first-mover advantage, unless it can follow up with rapid expansion. Second, we employ an analysis of the computing requirements for the development of current generative AI models. Our goal is to show that the concentration of computing power in the hands of a few players is directly linked to concerns of contestability and fairness as defined in the DMA. Third, data resources represent a powerful input that is not available for all. Indeed, large and integrated companies developing generative AI models benefit from troves of data unavailable to smaller actors. Finally, we argue that a variety of companies are opting for integrated services, combining search, email, personal assistants, and many others, with generative AI models through a user interface. We link this phenomenon with contestability issues as defined by the DMA.(iii) Generative AI as a core platform service: The DMA has the potential to address gatekeeping concerns in the generative AI industry if it is amended to include generative AI as a core platform service. Our preliminary conclusion is that the DMA’s definition of generative AI as a core platform service can be tied to the AI Act’s definition of general purpose AI systems. This category was added to the AI Act with generative AI applications in mind following the advent of large language models. This approach would be in keeping with the DMA’s definition of core platform services in reference to other EU legislation.(iv) Obligations for generative AI gatekeepers: Some of the existing gatekeeping obligations would apply to generative AI gatekeepers, should the DMA be amended to include generative AI as a separate core platform service. For example, a generative AI gatekeeper would be prevented from combining personal data across its services (Article 5(2)). However, our hypothesis is that certain contestability and fairness issues that the industry currently exhibits may not be addressed with existing obligations under Articles 5 and 6 of the DMA. We are currently investigating whether novel obligations should be added to the DMA for generative AI gatekeepers.