Entre octobre 2019 et juin 2020, 150 citoyens tirés au sort se sont réunis pour définir des mesures de réduction d’au moins 40 % des émissions de gaz à effet de serre d’ici 2030 (par rapport à 1990) dans un esprit de justice sociale. Leurs propositions ont contribué à la préparation de la loi « Climat et Résilience ». Une trentaine de chercheurs issus de différents milieux (thématiques, méthodologiques, géographiques) ont observé les travaux des membres de la Convention. Cet article propose une synthèse descriptive de ces travaux, ainsi que les premiers résultats sur l’apport démocratique de cet exercice inédit en France. Abstract Launched in 2019, the French Citizens’ Convention for Climate (CCC) tasked 150 randomly chosen citizens with proposing fair and effective measures to fight climate change. This was to be fulfilled through an “innovative co-construction procedure”, involving some unspecified external input alongside that from the citizens. Did inputs from the steering bodies undermine the citizens’ accountability for the output? Did co-construction help the output resonate with the general public, as is expected from a citizens’ assembly? To answer these questions, we build on our unique experience in observing the CCC proceedings and documenting them with qualitative and quantitative data. We find that the steering bodies’ input, albeit significant, did not impair the citizens’ agency, creativity, and freedom of choice. While succeeding in creating consensus among the citizens who were involved, this co-constructive approach, however, failed to generate significant support among the broader public. These results call for a strengthening of the commitment structure that determines how follow-up on the proposals from a citizens’ assembly should be conducted.