A new review suggests that artificial intelligence is increasingly effective at identifying stigmatizing language in healthcare-related discussions, but there is still limited evidence that it can successfully reduce stigma in real-world settings.

Researchers analyzed 70 studies published between 2016 and 2025 that examined the relationship between AI and health-related stigma. Their findings, published in the journal npj Digital Medicine, show that most research has focused on using AI to detect and measure stigma rather than actively address it.

Health-related stigma remains a major obstacle to equitable healthcare. As AI becomes more deeply integrated into health communication, prevention, and treatment, experts are paying closer attention to whether these technologies help reduce discrimination or inadvertently reinforce it.

The review found that machine learning and natural language processing tools are commonly used to identify stigmatizing language across large digital platforms, including X, Reddit, Weibo, and Facebook. These systems can analyze vast amounts of online content and uncover patterns of prejudice, stereotypes, and discrimination.

However, the researchers also found evidence that AI systems can perpetuate stigma. Some studies showed that language and image-generation models reproduced negative stereotypes associated with mental illness, disability, and other health conditions. Concerns about bias, data misuse, and misrepresentation have become increasingly important with the rise of large language models.

The review identified four main areas of research: AI measuring stigma, stigma influencing AI adoption, AI contributing to stigma, and AI helping to reduce stigma. Of the 70 studies reviewed, 42 focused on detecting stigma, while only four examined interventions designed to reduce it.

Those limited studies suggested that conversational AI agents may help decrease stigmatizing attitudes, particularly when they share first-person narratives about living with a health condition. Researchers noted, however, that these findings are based largely on small experimental studies and have yet to be confirmed in long-term clinical or community settings.

The authors conclude that AI currently serves more as a tool for analyzing stigma than for combating it. They also highlight the need for more diverse research, broader representation of health conditions, greater cross-cultural perspectives, and more real-world evaluations of AI-driven anti-stigma interventions.