Background: Public involvement activities (PIAs) may contribute to the governance of ethically challenging biomedical research and innovation by informing, consulting with and engaging the public in developments and decision-making processes. For PIAs to capture a population's preferences (e.g. on issues in whole genome sequencing, biobanks or genome editing), a central methodological requirement is to involve a sufficiently representative subgroup of the general public. While the existing literature focusses on theoretical and normative aspects of ‘representation', this study assesses empirically how such considerations are implemented in practice. It evaluates how PIA reports describe representation objectives, the recruitment process and levels of representation achieved. Methods: PIA reports were included from a systematic literature search if they directly reported a PIA conducted in a relevant discipline such as genomics, biobanks, biotechnology or others. PIA reports were analyzed with thematic text analysis. The text analysis was guided by an assessment matrix based on PIA-specific guidelines and frameworks. Results: We included 46 relevant reports, most focusing on issues in genomics. 27 reports (59%) explicitly described representation objectives, though mostly without adjusting eligibility criteria and recruiting methods to the specific objective. 11 reports (24%) explicitly reported to have achieved the intended representation; the rest either reported failure or were silent on this issue. Conclusion: Representation of study samples in PIAs in biomedical research and innovation is currently not reported systematically. Improved reporting on representation would not only improve the validity and value of PIAs, but could also contribute to PIA results being used more often in relevant policy and decision-making processes.