Bringing the public voice into frontline research is an NIHR priority. Researchers regularly acknowledge that PPIE input improves both work and outcomes. Public participants are keen to influence the direction of research. In the AI space, with its issues of contested data, trust and transparency, establishing a meaningful dialogue between different constituencies takes on special importance. So why is PPIE often bland, ineffective or just goes badly wrong? This talk was presented as part of the AIM RSF Open Invitation Seminar.