Chronic Obstructive Pulmonary Disease (COPD) is the most common chronic respiratory disease in the world. Fluctuations in symptoms frequently occur in COPD and might require additional treatments. Automatic monitoring or diagnosis of these fluctuations may aid the management of people with COPD. Speech analysis is a possible remote non-invasive methods to achieve this. However, this analysis is disease and possibly language specific and the development of speech models requires high-quality and high-quantity data, which are often not available. In this study we introduce SPEAKtoCOPD: a dataset of Dutch speech collected with a flash mob study targeting people with a respiratory disease (specifically COPD). In this paper, we describe the flash mob methodology, our collected data, and perform an automatic quality check to assess whether the speech tasks have been performed correctly. We aim to publish our gathered data for researchers to improve speech analysis for respiratory health.