Kaposi's sarcoma in individuals living with HIV: comparative assessment of AI-based clinical responses using a standardized questionnaire set

Comprehensive Summary

This article aimed to test the capability of common digital AI platforms in accurately answering standardized questions regarding HIV-related Kaposi’s sarcoma. A list of 10 standardized questions was created and entered into each of 3 platforms, ChatGPT, Consensus, and DeepSeek. ChatGPT and Consensus were both able to answer all questions correctly, while for one question DeepSeek had a technical error. Consensus also provided citations, and ChatGPT provided visual support and utilized algorithms for replicating clinical decision-making. The key finding was that all three models had comparable quantitative competency, but were unique in qualitative output quality. If these AI models were to be used by physicians, while they can provide objectively correct information a notable proportion of the time, they should not be used in place of standard medical treatment determination practices.

Outcomes and Implications

This research is important as it assesses the capabilities of commonly used AI models in providing medical counsel. This is important as the use of these models to give clinical advice is growing in incidence, and it should be ensured that users are not misled about their conditions. Toplu notes that AI models can be used to assist physicians in treating complex compound conditions such as HIV-related Kaposi’s sarcoma, but should be used alongside clinical guidelines and medically accurate resources rather than alone.