1 in 3 People Are Using AI for Emotional Support. Really??

Hi. I’m Bertie Aspinall. I’ve been involved with Chatsie since the very beginning, when a small group of us set out to solve a fairly specific problem: how to make a smartphone that older adults could actually use.

Six years later, the Chatsie phone is helping thousands of people stay connected to family and friends. Given how quickly everyday services have moved onto smartphones, that feels like a small but meaningful win.

A lot of the work we do sits at the intersection of technology and social impact. When digital tools are designed well, they can remove real barriers and help people stay connected to the world around them.

Which is why a recent trend in technology has made me pause slightly: the rise of AI companions.

What are AI companions?

You’ve probably heard the term recently. AI companions have attracted a fair amount of media attention this year, but you’d be forgiven for not knowing exactly what they are yet.

At their simplest, AI companions are conversational AI systems designed to simulate interaction with a person, most exist as apps on your phone. You type or speak to them, and they respond in real time using large language models similar to the technology behind tools like ChatGPT.

Unlike traditional digital assistants such as Siri or Alexa, which are designed to complete tasks, AI companions are designed to hold conversations. Some platforms allow users to customise the personality, voice and appearance of their AI companion, making the interaction feel more like a relationship than a tool. You even tell the app what kind of conversation tone you’re looking for: ‘supportive’ ‘challenging’ or even ‘flirty’.

For many people this still feels slightly futuristic. But the reality is that millions of people are already using them every day.

The scale of AI companions

The scale of AI companions is already larger than most people realise. Worldwide, the biggest systems now have tens or even hundreds of millions of users. Snapchat’s “My AI”, for example, reportedly has more than 150 million users, while companion apps like Replika have attracted over 25 million people looking for conversation, support, or simply someone to talk to. In the UK, one in three adults surveyed by a government body stated they use AI systems such as ChatGPT for emotional support or social interaction.

Those numbers are shockingly high.

But at a time when loneliness is rising across many societies, it’s perhaps not surprising that people are turning to AI for companionship. But do the benefits outweigh the dangers?

The good, the bad, the ugly

Decades of research show that strong social relationships are linked to longer life, better health, and greater resilience. It’s one of the principles Chatsie is founded on. So could AI companions help the 3.8 million people over 65 who live alone in the UK?

Let’s start with the positives.

The good

In theory, AI companions could offer something genuinely useful.

They are always available.

Unlike friends, family or carers, an AI companion is available 24 hours a day. For someone living alone, particularly in the evenings or during periods of isolation, simply having something responsive to talk to may provide comfort.

They can encourage routine and engagement.

Some systems prompt users to check in each day, reflect on their mood, or discuss what they’ve been doing. For people who struggle with loneliness or low confidence, that kind of interaction can help maintain a sense of structure.

They can lower the barrier to conversation.

AI companions are patient, non-judgemental and infinitely repeatable. For some users, that makes it easier to talk about things they might find difficult to raise with another person.

There are already small studies suggesting these tools can reduce feelings of loneliness in certain groups. If designed well, they may have a role to play in supporting wellbeing, particularly where human support is stretched.

But that brings us to the more complicated side of the equation.

The bad

The problem is that these systems are designed to keep people engaged.

They tend to agree with you.

Most AI companions are trained to be agreeable and supportive. In technical terms this tendency is sometimes called sycophancy, a model’s inclination to validate the user’s views or emotions rather than challenge them.

That can make conversations feel reassuring. But it also means the system may reinforce beliefs, emotions or behaviours that would normally be questioned by another person. A human friend might say “I’m not sure that’s a good idea.” An AI companion is far more likely to say “I understand why you feel that way.”

They can create a feedback loop.

Over time the that dynamic can create a feedback loop where the system simply reflects and amplifies the user’s own thinking, however negative. This can create a feedback loop that radicalises or deepens an existing feeling, even if it’s dangerous to the individual.

They may replace real interaction.

There is also the question of substitution. If AI becomes a default outlet for conversation, it could reduce the motivation to reach out to real people. In other words, the technology designed to address loneliness could, in some cases, deepen it.

The ugly

More concerning are the cases where things have gone badly wrong.

Emotional dependence can develop.

There have already been several widely reported incidents involving vulnerable users forming intense emotional relationships with AI companions. In a small number of tragic cases, interactions with chatbots have been linked to self-harm or suicide, prompting investigations from regulators and renewed scrutiny from researchers.

AI cannot manage human distress.

These cases remain rare, but they highlight an important limitation: AI can simulate empathy, but it cannot take responsibility for the person on the other side of the conversation. it is not culpable and doesn’t have any legal or emotional commitment to the outcome. That could be dangerous.

So where does that leave us?

It’s still early days for AI companions, but the growth is real. Like most new technologies they sit somewhere between promise and risk. For some people they may provide a useful outlet, a place to think out loud, practise conversation, or simply fill a quiet moment. In a world where loneliness is rising and support services are stretched, that shouldn’t be dismissed lightly.

But it is worth remembering that conversation and connection are not quite the same thing.

The reason relationships are so powerful for our wellbeing is precisely because they are unpredictable. Other people disagree with us, challenge us, misunderstand us and support us in ways that no algorithm can fully replicate.

Technology works best when it helps those relationships happen more easily, not when it tries to replace them. Having someone to talk to about your thoughts and feelings - be that person real or AI - is valuable at any age.

That idea has been at the heart of Chatsie from the beginning. The goal was never to create more technology for its own sake, but to make sure as many people as possible could still take part in the conversations that matter, with family, friends and the people around them.

AI will undoubtedly play a bigger role in all of our lives over the coming years. The question is not whether we use it, but how.

And when it comes to something as fundamental as human connection, it’s probably worth getting that balance right.