Back

Using novel internet-based sources for Population Size Estimation of Gay, bisexual, and other cisgender men who have sex with men

Mi, Y.; Garner, A.; Niang, M.; Hong, C.; Howell, S.; Herman, J.; Holloway, I.; Baral, S.

2026-09-16 hiv aids
10.64898/2026.09.14.26363087 medRxiv
Show abstract

Population size estimates (PSE) for gay, bisexual, and other cisgender men who have sex with men (GBMSM) remain difficult where stigma and criminalization limit data collection and HIV resource allocation. We aim to assess whether Google Trends and a gay social networking app can provide standardized, low-cost population size estimates (PSE) for gay, bisexual, and other cisgender men who have sex with men (GBMSM) in Senegal, Ghana, Mozambique, and Indonesia. We analyzed country-level and local GT relative search volumes for "porn" and "gay porn" from September 2024 to September 2025. Pornhub global review data were applied to produce proportional GBMSM PSEs. App estimates were derived from monthly active users and extrapolated to annual users. Proportional estimates were multiplied by the male population aged 15-49 years to obtain absolute PSE and compared with UNAIDS Key Population Atlas estimates. Sensitivity analyses used survey-based multipliers and alternative search-term combinations. GTPSE ranged from 0.37% in Mozambique to 1.92% in Indonesia, while app-based estimates ranged from 0.29% in Mozambique to 1.34% in Ghana. GTPSE exceeded UNAIDS estimates in Ghana, Mozambique, and Indonesia. App-based estimates exceeded UNAIDS estimates in Ghana and Mozambique. Maputos estimated GBMSM proportion was substantially higher than Mozambiques national estimate. GTPSE generally exceeded UNAIDS estimates, which may reflect underestimation in stigmatizing settings where GBMSM are less likely to be reached through field-based methods or mainstream health services. Social network app estimates were lower in some countries, highlighting variability in platform access and geographic differences in the concentration of social networks. Internet-derived data may offer a low-cost, feasible approach to GBMSM PSE in low- and middle-income settings. Further validation through digital data linkages could strengthen HIV prevention and treatment planning.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.