In order to achieve large scale behaviour change, persuasive technology should look beyond the minority of users who already behave sustainably and aim to involve larger segments who are willing to behave more sustainably, but are not doing so yet. There is strong consensus that effective persuasive agents should be personalised; people tend to be more persuaded by messages that are relevant to their circumstances and that are delivered in a way that is relatable to them. Several open questions remain in relation to what the optimal persuasive language looks like for different users, and whether an agent is more persuasive if it speaks like its different users. Through a series of studies, this research proposal sets out to develop a persuasive agent with personalised dialogue to help people behave more sustainably. We specifically explore how different people write persuasively, and look for possible user-chatbot similarity effects to enhance persuasion. Preliminary results which isolated personality markers in persuasive language are presented, followed by a proposed methodology for subsequent studies.