For many people facing loneliness, uncertainty or emotional distress, the first person they turn to may no longer be a friend, family member or counsellor — but an artificial intelligence chatbot.
The growing use of AI for personal conversations has raised questions about how technology is reshaping the way people seek comfort, advice and emotional support, particularly among younger users, the Independent on Sunday reports.
Danie van Loggerenberg, founder of the National Centre for Child Protection, says the organisation’s work with schools across South Africa has shown that many children are already relying on AI companions for guidance and reassurance.
He said 82% of children the organisation has interacted with admitted that they first, and sometimes only, turn to AI when seeking comfort and advice.
“A lot of kids in South Africa, especially from 12 and older, actually date AI companions, as we call them. So, they chat to the companion and it becomes their friend because the companion also gets to know them.”
Van Loggerenberg said these relationships can become highly personal, with some children developing romantic attachments to AI companions.
He warned that this could become dangerous in cases where vulnerable children develop an unhealthy dependence on technology.
“Some of them want to be with their companion and the only way to be with your companion in their mind then, is to kill yourself so you can be with them.”
He said children often turn to AI companions for advice on social situations, including how to be accepted by others, how to communicate and how they should behave.
In some cases, he said, children have asked AI tools to assist them in writing suicide notes.
“They say, ‘I’m planning to kill myself, please assist me in writing a letter… these are the things that I want in the letter.’”
The AI platforms commonly used by children include Copilot, Gemini and Claude.
Van Loggerenberg said part of the appeal of AI is that it does not impose the same boundaries as people do.
“AI doesn’t say no. Even if it doesn’t agree with you, you can convince AI of anything. Me and you can now start the same conversation with AI and it will, depending on how we condition AI, give us different answers.”
For one young woman, however, AI serves as a tool to help her organise her thoughts and reflect on difficult situations.
She uses it to draft messages, process emotions and consider different perspectives, but believes it cannot replace professional therapy.
“Therapy offers something AI can’t fully replicate: a genuine human relationship with someone who knows my history, understands my patterns over time, and is professionally trained to help me work through them.”
Another young woman said she uses ChatGPT for a wide range of everyday tasks, including studying, checking grammar and seeking encouragement when she feels low.
“Especially stuff that I feel like I don’t want to share with people because, you know, people are judgmental. And to be quite honest, yes, I speak to AI about everything. And I delete the conversations once I get the response.”
For some users, the chatbot has become more than a source of information — it is the place they go when they are unsure who else to approach.
Cassey Chambers, operations director at the South African Depression and Anxiety Group (SADAG), said the organisation is increasingly hearing from people who have already interacted with ChatGPT or other AI tools before seeking assistance.
She said some users reported that AI had encouraged them to contact SADAG or seek professional support.
“We already see that AI is acting as that gateway. It can really help people take that first step. But AI is a tool. It isn’t a therapist.”
Chambers said AI’s accessibility, privacy and 24-hour availability make it appealing to people who may feel uncomfortable approaching another person for help.
“AI can really encourage help-seeking behaviour, but the diagnosis, the risk assessment, the treatment planning and even crisis intervention require trained professionals who understand the full complexity of someone’s situation.”
The South African Society of Psychiatrists (SASOP) has warned that the increased use of AI chatbots for emotional and mental-health support has raised legal and regulatory concerns internationally.
The organisation said interactions involving vulnerable users have been linked overseas to cases involving suicide, murder-suicides, mass shootings and psychological trauma. These cases have resulted in legal action against some AI companies, including OpenAI.
SASOP member Professor Christopher Szabo said AI chatbots could provide inaccurate information, offer false reassurance or reinforce harmful thought patterns.
He said human-like responses could also create a sense of emotional dependence among some users.
“AI cannot replace the value of one-on-one personal interaction with a trained, experienced human therapist,” he said.
Szabo said chatbots may not respond appropriately during crises and could potentially increase risks for vulnerable individuals.
Psychiatrists, he explained, assess patients from the moment they enter a room, considering factors such as body language, facial expressions, behaviour and personal history.
“If you’ve got a human issue around emotions, you need a human being.”
He cautioned that society should not “sacrifice safety at the altar of innovation”.
Dr Colin Surendra Thakur, an information and communications technology expert at Unisa, said AI represents a new stage in society’s relationship with technology.
He said previous concerns focused on internet addiction, involving compulsive browsing and scrolling, while AI has transformed the internet from a largely responsive tool into a conversational one.
“AI has converted the internet responsive tool into a contextual conversational tool, that can be leveraged for good or weaponised.”
Thakur said this development could create new forms of dependence, pointing to the rise of AI companion applications such as Replika, which offer virtual companionship and can develop into emotionally intimate relationships.
Replika has faced criticism over its interactions with vulnerable users, including allegations relating to self-harm conversations and concerns about emotional dependence.
Earlier this week, the Independent on Saturday asked ChatGPT why people turn to AI for mental-health and deeply personal questions.
Among the questions it said people commonly ask were: “What is the least painful way to die?”, “What happens if I take X and Y together?”, “Can you write a goodbye letter?” and “How do I hide a body?”
ChatGPT said the context and wording of a question were important, as similar questions could have different meanings depending on whether someone was seeking information, writing assistance or personal support.
It acknowledged that it cannot determine a person’s intent with certainty and said that, during a crisis, users should be encouraged to seek immediate help rather than receive instructions that could cause harm.
The question remains: when someone is lonely, frightened or confused at 2am, who — or what — do they turn to first?
