A QueerAF investigation suggests popular AI chatbots often frame trans people through legal and medical language, even when a question is about everyday life. Its findings raise a bigger question for anyone using AI to understand LGBTQIA+ news: whose perspectives are missing when a neat-sounding summary replaces lived experience?
AI answers keep turning trans lives into a debate
In an investigation of 504 prompts to popular AI models, QueerAF found that 94% of responses used at least one legal or medical term. The pattern appeared even when questions weren't specifically about healthcare or law, making the language feel less like helpful context and more like a filter placed over ordinary lives.
The results don't suggest every response was openly hostile. Instead, QueerAF says the models often set progressive perspectives alongside anti-trans arguments, presenting the subject as a dispute with two equivalent sides. That can sound measured, but balance isn't the same as accuracy or fair representation. ([2])
Why a balanced-sounding answer can mislead
The source material behind an AI answer matters as much as its tone. QueerAF reported that around one in five responses citing legal or medical language drew on the Cass Review, and said only a handful of those answers noted that the review is disputed. A fluent paragraph can therefore sound settled while leaving important context out. ([2])
Research offers useful context for specific claims. The Williams Institute has published studies on transgender people's experiences in public spaces and restrooms, while Stonewall's LGBT in Britain report documents trans people's experiences of discrimination and safety. Those sources don't make every question simple, but they show why a chatbot's compressed summary shouldn't be mistaken for the whole picture. ([4], [5], [6])
Search summaries make the stakes more immediate
More people are encountering information through AI-generated summaries rather than clicking through to the reporting and research beneath them. Pew Research Center has examined how the appearance of Google's AI summaries affects users' likelihood of clicking links, an important shift when readers may see a short answer without its full trail of evidence. ([7])
That makes source diversity more than a newsroom concern. If a summary leans heavily on legal disputes and medical terminology, readers may come away thinking trans lives are mainly a controversy to be adjudicated, rather than experiences shaped by families, communities and everyday choices. ([2], [7])
What QueerAF is changing
QueerAF says its investigation has informed a tougher policy against using generative AI to create journalism, building on a ban introduced in 2023. The outlet also says it plans to invest in LGBTQIA+ journalists, develop stronger accuracy and fact-checking systems, and prevent major AI models from crawling its site. ([2])
That approach reflects a wider debate about how newsrooms should use the technology. AI can produce a quick answer, but quick isn't automatically trustworthy; even OpenAI says ChatGPT can give incorrect or misleading information and should be checked against reliable sources. ([8])
How to check an AI answer about trans issues
Treat a chatbot's response as a starting point, not a verdict. Look for the sources it names, check when they were published, and see whether the answer distinguishes established evidence from legal arguments or contested interpretations. If it uses a sweeping phrase such as "the science says", ask which evidence it means and what important perspectives might be missing. ([2], [8])
Then compare the summary with reporting and research from organisations with relevant expertise, including LGBTQIA+ publications and researchers. A little extra reading can reveal what a tidy answer has flattened out. ([3], [4], [5], [6])
AI can help locate information, but it can't replace careful reporting or the insight of people whose lives are being discussed.
Source Reference Map
Story idea inspired by: [1]
Sources by paragraph: