Anxiety responses and testing intentions among gay and bisexual men using an AI-powered HIV/STI risk assessment tool: a quasi-experimental study

Introduction Web-based tools for the assessment of the risk of sexually transmitted infections (STI) may increase testing but also may induce anxiety. We evaluated anxiety responses and testing intentions among gay, bisexual and other men who have sex with men (GBMSM) using MySTIRisk , an artificial intelligence-powered STI risk assessment tool, compared to a standard sexual health information webpage for GBMSMs. Methods We conducted a quasi-experimental pre-post study at the Melbourne Sexual Health Centre between April and October 2024. Participants were allocated to either MySTIRisk , which provides personalised HIV/STI risk assessment, or a standard webpage providing general sexual health information for GBMSM on alternating days. We measured anxiety using the State-Trait Anxiety Inventory (STAI-6), which ranges from 6 to 24, before and after participants viewed their assigned websites. We defined clinically significant anxiety changes as ≥ 3 points on STAI-6. We used multivariable ordinal logistic regression analysis to evaluate clinically significant anxiety changes while controlling for baseline scores and demographic variables. Results Our study population had a median age of 34 years (IQR: 28–42); 48% were born in Australia, and 69.7% had tertiary education. The baseline characteristics were similar between MySTIRisk ( n = 150) and control ( n = 150) groups. In the MySTIRisk group, STAI-6 scores increased significantly from baseline (median [IQR]: 10 [8, 12]) to post-intervention (11 [8, 13], P = 0.001). The control group showed a decrease from baseline (11 [9, 12]) to post-intervention (10 [8, 13], P = 0.01). Despite increased anxiety, MySTIRisk users maintained high user acceptability (92.0%) with similar testing intentions between groups (87.3% vs. 82.7%, P = 0.6). The multivariable ordinal logistic regression analysis showed that the intervention’s effect on anxiety was independent of demographic characteristics, except for employment status. Conclusions While MySTIRisk increased anxiety in some users, it maintained high acceptability and did not deter testing intentions. These findings support implementing such tools with appropriate anxiety management strategies.