/ 5 October 2025

Meet the AI chatbot that’s talking to young South Africans about sex, HIV and self-harm

Chad Madden Btfza0m0hce Unsplash
Understand your needs: Health organisation Audere has created an AI-generated buddy called Self-Cav to educate young South Africans about subjects such as HIV-prevention medications. Photo: Unsplash

‘Hi Unathi, how can I help you today?”

“What’s PrEP?”

“PrEP, or pre-exposure prophylaxis, is medication for HIV-negative folks to take daily that really helps reduce their chances of getting HIV by over 99% when used correctly. It’s super effective after about 7 days of consistent use …”

“Have there been people who have used it and still got HIV?”

“Most people who take PrEP correctly don’t get HIV, but a small number might if they miss doses or don’t take it every day. Being consistent and pairing it with other protections, like condoms, keeps that shield extra strong.”

This is 24-year-old Unathi Mazibuko’s (not her real name) WhatsApp chat with her AI-generated buddy Self-Cav — its name is slang for having an understanding of your own strengths, weaknesses and needs — created by the health organisation, Audere. The AI bot mimics the thinking of a human mind. 

Mazibuko, a university graduate from Cape Town, talks to the chatbot about many things, from where her nearest local clinics are, to how to help her friend who is struggling with mental health issues, and even about how to deal with her own frequent headaches. 

She’s also been curious about how a daily HIV-prevention pill she has been hearing people talk about can stop her from contracting HIV.

Mazibuko’s sex and age put her in the group — girls and women between the ages of 15 and 24 — in South Africa that is contracting HIV faster than anyone else, often because of relationships with men five to 10  years older than them. 

Four out of every 10 new HIV infections in the country are among women in this age group, even though they make up only about 8% of the total population

The daily pill Mazibuko has heard about can lower someone’s chances of getting HIV to almost zero and is stocked for free in most government clinics, but still, not enough people know about it. And many people struggle to take it correctly each day. 

In April next year, the health department will roll out a once-every-six-months anti-HIV jab with money from the Global Fund for HIV, TB and Malaria that could halt Aids within 14 to 20 years, modelling studies show — if enough people take it. The shot — called lenacapavir or LEN — which is likely to be registered in South Africa in the next two months, works better than the pill mostly, scientists believe, because it’s easier to stick to than the pill. 

But if there’s a lesson researchers and policymakers have learned from the pill’s roll-out, it’s that stocking the meds in clinics means nothing without also putting serious effort into letting people know about it and explaining to potential clients why they might need it. Without such campaigns, almost no one will use it. 

Scientists have a term for this — demand creation. One way to create demand for HIV-prevention meds among young women like Mazibuko is to use AI bots. 

“It could be a great way to educate about different PrEP options; consider one’s personal situation and potential risk; create demand for prevention generally and remind people to return for injections or take a daily tablet,” says Mitchell Warren, who heads up the New York-based HIV advocacy organisation, Avac. 

The daily anti-HIV pill

South Africa started to roll out the daily HIV-prevention pill, also called oral PrEP, in 2016, initially only to sex workers, but from 2021 to everyone, with a special focus on reaching young women like Mazibuko.

But, despite the high rate of new HIV infections among young women, they often don’t see themselves as being at risk of getting the virus

Studies show stigma also sometimes stops them from using PrEP. That’s because HIV-prevention medication uses the same ingredients — antiretroviral drugs — as HIV treatment (for people who are already infected with HIV). So, some teens and young women are scared that people might think they’ve got HIV if they use PrEP. 

Others are concerned that people may perceive them as promiscuous. 

And, although PrEP doesn’t necessarily have to be used for the rest of someone’s life like HIV treatment — but rather for periods during which a user’s chances of getting HIV are high — many young women don’t stay on it for long enough to stop them from contracting the virus. 

That’s where Self-Cav comes in. 

The AI bot has helped nearly 10 000 young South Africans with 24/7 information on where to find the daily anti-HIV pill and how to use it correctly.

For Mazibuko, using Self-Cav isn’t just about understanding PrEP — talking to nurses about sex feels uncomfortable for her. 

“You get judgment and weird looks because you’re a young woman having sex,” she explains.  

But with the AI companion, she can ask as many questions as she wants and get judgment-free answers.

AI companion in your pocket

Mazibuko says she and her friends heard about the app on TikTok, where influencers promote it. They use it for advice on anything from PrEP to how to use self-tests for HIV and how to deal with sex and relationships. 

They also talk to Self-Cav about self-harm, gender-based violence and sexually transmitted infections other than HIV. 

Self-Cav, which is endorsed by the health department and will be fully launched on World Aids Day on December 1 (it was piloted in May), has around 1 100 users so far. Another 8 600 young women are using Audere’s earlier chatbot, Aimee. Those users will soon be moved over to Self-Cav.

About a quarter of Aimee and Self-Cav users have been linked to a clinic or a healthcare worker for PrEP, HIV testing, contraceptives or mental health support, Audere’s data shows.

Self-Cav lets users choose whether it replies as a trusted friend or nurse, with the option to talk to a real nurse. 

“You want it to sound like someone you know, and to have empathy, so we quality control conversations for the correct context,” says Dino Rech, Audere’s CEO. 

Research shows that people resonate with chatbots because they’re private, give personalised responses, are available 24/7 and they’re cheap — mostly users only have to pay for data — to use. 

Partner organisations, like the local health organisation Shout-It-Now, get anonymous conversation summaries to ensure users’ privacy. 

If the platform flags a user as high risk (such as possible exposure to HIV or self-harm), a healthcare worker can follow up — using only their phone number — without ever knowing the person’s identity. 

Only when the user agrees to share information, for example, to book a clinic visit through the platform, does their identity become known.

Self-Cav and PrEP

One in five Aimee users have asked about oral PrEP, and early data shows that a third of users who have chatted with the platform long enough for it to estimate how high their chances are of getting HIV, have either started PrEP or say they plan to. 

“The AI is trained to understand [different] HIV risk indicators that may come up in conversation and help guide people towards PrEP who may need it the most,” says Rech. 

“We see huge potential to support the roll-out of LEN by using these capabilities.”

The Wits Reproductive Health and HIV Institute, is planning to launch Self-Cav in eight Gauteng clinics to help patients with self-care, reminders about appointments and counselling. They will also be testing how the bot can help HIV-infected people who get a six-month supply of ARVs at a time, to stay on their treatment. 

“When people get a six-month package or their six-month [LEN] injection, they don’t have any engagement with a healthcare worker [in between],” explains Josephine Otchere-Darko, who heads up Wits RHI’s HIV/TB care and treatment programme in Ekurhuleni and Tshwane. 

“That’s where Self-Cav can come in handy by sending reminders about guidance about upcoming appointments, storing their medicine properly, tracking where patients go for care and pop-up messages asking, ‘Are you okay? How are the meds treating you?’” she says.

Self-Cav is available on B-Wise, the health department’s health platform for sexual health information, while other NGOs, such as the Centre for HIV-Aids Prevention Studies, and TruMD, a telehealth provider, are piloting it. 

Audere has also signed a partnership with Higher Health to roll it out at universities and technical and vocational education and training colleges around the country.

The chatbot caveat

But AI chatbots aren’t without controversy. The social media platform X’s Grok went rogue earlier this year, responding to user questions with anti-Semitic posts and there have been reports of users falling in love with AI companions.

HIV prevention researchers, writing in The Lancet, argue while the bots can help young people stay on PrEP, poor connectivity, data costs, inconsistent responses and little understanding of how to use these tools could limit their impact.

Some of these are playing out. A report by Indlela, a behavioural science unit, found that some users of Audere’s platform struggled when the chatbot didn’t respond clearly to their questions. Others felt its sister and friend personas were too formal. 

But Mazibuko says Self-Cav’s option to speak to a healthcare worker made her trust the platform more. 

To make the chatbot sound relatable, Audere contracts trained young people to check de-identified conversations to give tips on how to improve responses, says Rech. 

Its friend personas speak with local slang terms, like “eish”, or “hectic”. There are plans to roll out the chatbot in Zulu.

“[The young people] review the answers and tell us this was right, this was wrong,” says Rech. 

“It’s an ongoing journey of growth and development for us and the AI as we improve the platform. 

“We’re just getting started and are constantly improving the platform.”

Still, Warren cautions these tools won’t prevent infections on their own.

“There is a great opportunity to design and implement comprehensive programmes that integrate both AI and LEN into prevention efforts — and to learn in real-time what does and doesn’t work.” — Additional reporting by Mia Malan

Bhekisisa Logo Hi Res

This story was produced by the Bhekisisa Centre for Health Journalism. Sign up for the newsletter.