detrans.ai: A Counter Narrative | Peter James Steven

Peter James Steven built detrans.ai to give detransitioners a voice that mainstream AI models have largely suppressed. Drawing on a growing database of first-hand accounts, the chatbot surfaces statistical patterns in why people transition and later reverse course. At a moment when the Cass Review has forced UK institutions to confront the evidence gap around long-term outcomes and patient regret, tools that centre detransitioner testimony add an important dimension to the public record.

Peter James Steven is the developer behind detrans.ai, an AI chatbot built specifically to aggregate and surface the experiences of people who have detransitioned — those who underwent medical or social gender transition and subsequently reversed course. The project grew out of personal connections to detransition stories and a concern that dominant AI systems were trained on data reflecting affirmation-only perspectives, giving little weight to the growing body of accounts from people who came to regret transition. In conversation with the Beyond Gender hosts, Steven walks through the technical choices behind the project, including the selection of underlying language models and the considerable challenge of building a tool that handles sensitive material responsibly. He also discusses the statistical picture that has emerged from the database, covering patterns in the reasons people gave for transitioning in the first place and the range of circumstances that eventually led them to stop. The relevance of this work to the UK evidence landscape is direct. The Cass Review, published in April 2024, identified a significant gap in follow-up data for patients who had passed through gender identity services: long-term outcomes were poorly tracked and the voices of those who came to regret treatment were largely absent from the clinical literature. Detrans.ai represents an attempt, from outside formal research institutions, to begin filling that gap through systematic collection of lived experience. It is not a clinical data set, but it points toward a methodology for capturing testimony that the NHS and academic researchers have been slow to develop. The episode also addresses the broader problem of ideological capture within AI systems. When mainstream tools consistently reflect affirmation-based framings, users seeking information about detransition, regret, or non-affirming therapeutic approaches receive a distorted picture. Steven's argument is that a specialised, counter-narrative model can correct for that bias and offer something closer to the full range of human experience. The discussion touches on identity, sexuality, and the psychological complexity behind the decision to transition or detransition, including what the patterns in the database reveal about the relationship between gender distress and other underlying factors. These themes sit closely alongside the Cass Review's finding that the majority of young people referred to gender clinics presented with significant co-occurring mental health conditions and neurodivergence — circumstances that warranted thorough psychological assessment rather than a fast pathway to medical intervention. For anyone engaged with the evidence on gender medicine, a tool designed to centre detransitioner accounts in a systematic way is a development worth examining.

The dossier behind this episode

Keira Bell, Bell v Tavistock and detransition in the UK

Keira Bell became the public face of a legal challenge to the former Tavistock and Portman NHS Foundation Trust’s Gender Identity Development Service (GIDS). Bell had received puberty-suppressing medication as an adolescent, later took testosterone and underwent surgery, and subsequently detransitioned. Together with “Mrs A”, the mother of a teenage girl, she sought judicial review of the lawfulness of the consent arrangements used when children and young people were referred for puberty blockers.

Read the evidence dossier