How the calculator works

The calculator estimates one thing: the probability that the exposure you describe resulted in HIV. It multiplies a handful of published figures together. It is a model, so it simplifies. It can't see your partner's viral load, your biology, or details you leave out, and it never replaces a test.

1. Risk per act

The starting point is how often one act passes HIV on, assuming the other person has HIV and is not on effective treatment. These are CDC's estimates, expressed per 10,000 exposures.

CDC lists oral sex as "low" (0 to 4 per 10,000) without a single figure; we use 1 in 10,000. Blood reaching the eyes, mouth or broken skin uses the estimate from healthcare-worker exposures. When you pick more than one type, we combine them as independent chances, 1 − (1 − p₁)(1 − p₂)…, rather than adding them.

kissing, bite or spit, shared sex toys and fluids on unbroken skin are counted as negligible. CDC rates biting, spitting and sharing sex toys as negligible, and HIV doesn't pass through saliva or unbroken skin.

Needlestick details

The needlestick figure is an average. A study of healthcare workers (Cardo et al., 1997) found these features raised the odds of infection. If more than one applies we use the largest, because they usually occur together.

2. Adjustments to the per-act risk

Viral load is extremely high in the first weeks after someone acquires HIV, and studies have found transmission several times more likely during that period than later; we use a factor of 10. An undetectable viral load sets sexual risk to zero, in line with the evidence behind U=U. Treatment also lowers risk from shared needles, but by an amount that hasn't been measured, so we leave needle and blood risk unchanged. The pregnancy figures come from a study of couples in seven African countries (Thomson et al., 2018), which found higher per-act risk late in pregnancy and after birth; risk early in pregnancy is somewhat lower than the figure we use.

3. The chance the other person has HIV

This is often the biggest factor in the whole estimate. If you know they have HIV, it is 100%. If they tested negative and the result is up to date, it is 0%. A negative test is up to date only if it was taken after the window period following their last possible exposure, with no risk since. If you're not sure, we treat their status as unknown, because someone who was recently infected can test negative while being at their most infectious.

For a partner of unknown status we build the figure in four steps.

Where they're from, and their sex

We start with the share of adults aged 15 to 49 living with HIV in the country or region you choose (UNAIDS estimates via World Bank indicator SH.DYN.AIDS.ZS (adults 15-49), retrieved 2026-10-29). UNAIDS also reports what share of people with HIV are women, so we split the figure by sex. In South Africa, for example, that gives about 23% for women and12% for men; in the UK, 0.27% for men and 0.13% for women.

UNAIDS doesn't publish modelled estimates for some countries. For China, Japan, Poland, Russia, South Korea, Türkiye, Ukraine, United States we use approximate figures from national reporting; these are marked "approx." in the list.

Whether a male partner is circumcised

Trials in Africa found circumcision cut men's risk of getting HIV through vaginal sex by 50 to 60%, so circumcised men there are less likely to have HIV. We treat a circumcised man as 50% as likely to have HIV as an uncircumcised one. Because the country figure already mixes the two, how much the answer matters depends on how common circumcision is there (Morris et al. 2016, corrected 2016). In the UK, where about 20.7% of men are circumcised, a circumcised man's figure is 44% lower. In Kenya, where about 91.2% are, an uncircumcised man's figure is 1.8× higher.

This only changes the chance that he has HIV. If he does, being circumcised doesn't lower the risk of passing it on: a trial in Uganda found no reduction in transmission to female partners (Wawer et al. 2009). We don't apply it to men who have sex with men, where the evidence is inconclusive, or to community figures, since circumcision rates are national.

Their community

In some countries, official surveillance shows much higher or lower HIV rates in particular communities. This is driven by social and economic factors and by sexual networks, not biology. Where those figures exist, an optional question lets you use them in place of the country figure:

The US figures are people with diagnosed HIV, scaled up for the share who don't yet know their status. National figures for other US communities, such as Hispanic/Latino and Asian American people, aren't yet included. The UK figures are the most recent published by community, but they are about ten years old.

Groups with higher HIV rates

If you're a man and had sex with a man, he is by definition a man who has sex with men, so that figure applies without ticking anything. We raise the odds rather than multiplying the percentage, so in high-prevalence countries the result stays a valid probability. The UNAIDS figures measure new infections rather than existing ones, so they are an approximation here. When several of these apply, or a community figure is also chosen, we use whichever gives the highest result rather than stacking them, because the groups overlap.

Treatment

Most people with HIV in most countries are on treatment, and people with an undetectable viral load can't pass HIV on through sex. So for sexual exposures we only count people with HIV who aren't virally suppressed. We estimate that share from each country's treatment coverage (UNAIDS, via the World Bank), multiplied by the93% of people on treatment who are virally suppressed worldwide (UNAIDS 2024). For needle and blood exposures we count everyone with HIV, since U=U hasn't been established for those. Where treatment data isn't available, we make no adjustment, which errs on the side of a higher estimate. Key populations often have lower treatment coverage than the national average, so for them this step may lower the estimate a little too much.

4. More than one exposure

With the same partner, their status doesn't change between acts, so we apply the prevalence once: prevalence × (1 − (1 − p)ⁿ). With different partners of unknown status, each act is a separate chance: 1 − (1 − prevalence × p)ⁿ. Partners who are virally suppressed are included only for blood exposures. Risk accumulates with repetition but never reaches certainty.

5. PrEP and PEP

The PrEP figures for consistent use are CDC's (about 99% for sex, at least 74% for injection drug use). The lower adherence levels and the PEP timing gradient are our assumptions: PEP works better the sooner it starts, but there are no precise figures for each day. Missed PEP doses double the remaining risk.

6. Negative test results

A negative test lowers the estimate using Bayes' rule. The question is how likely the test was to miss an infection on the day it was taken. Before a test's window opens, it tells us nothing. Across the window we assume the chance of a miss falls steadily from 90% to 0.5%. After the window closes, a negative result is treated as conclusive. If you entered several negative tests, we use the most informative one.

The 90% starting point is deliberately cautious, so an early negative result lowers the estimate less than it might. PEP can delay detection, so tests taken during the 28-day course are not used.

7. Risk levels

The label next to your number follows fixed thresholds:

• Very low: under 0.01% (less than 1 in 10,000)
• Low: 0.01% to 0.1%
• Moderate: 0.1% to 1%
• High: 1% or more (1 in 100 or higher)