In the age of artificial intelligence, that friend can now -- for better or worse -- be a chatbot.
This is the question Sheridan Voysey explores in his recent Friendship Lab conference talk on AI companionship. As the founder of Friendship Lab, Sheridan has spent years thinking about the nature of adult friendship, and his AI talk brings that work into one of the most pressing questions of our technological age: what happens when machines become remarkably good at making us feel heard, understood and even loved?
Watch Sheridan Voysey’s Friendship Lab talk on AI companionship
The statistics are confronting. According to the research Sheridan discusses, around one in three people have used ChatGPT, Claude or another AI chatbot for companionship. At first, that might seem strange. Why would anyone choose a machine when they could have a real friend?
But perhaps that’s the wrong question. The more important question is: Why are so many of us lonely enough to need or want one?
Friendship Lab points to a growing friendship crisis. More than 20 per cent of people in the UK, US, Australia and elsewhere have no close friends, while 51 per cent say they find making new friends difficult.
That changes the way we should think about AI companionship.
We shouldn’t simply mock people who find comfort in a chatbot. Loneliness is real. So is grief, social anxiety, moving to a new city, retirement, redundancy, divorce and all the other moments when the shape of our lives changes and the people who once surrounded us aren’t necessarily as close as they once were.
And AI is remarkably good at filling a gap.
Sheridan has written about discovering this himself when he asked ChatGPT to summarise one of his articles and received something he hadn’t expected: a compliment. The chatbot didn’t simply complete the task. It told him his writing was beautiful, moving and evocative. He describes it as his first AI affirmation.
And who doesn’t like being affirmed? But friendship has never really been about affirmation.
Sometimes a friend tells us we’re wrong. Sometimes they’re inconvenient and that’s part of the point.
A real friend has a life of their own. They have work, family, worries, commitments and bad days. They can’t always answer immediately and they certainly can’t be expected to exist entirely for us.
An AI can.
Sheridan puts it rather bluntly: if a companion is available 24/7 and exists entirely to serve your needs, perhaps it isn’t really a friend at all. It’s a servant.
That distinction gets to the heart of the matter. Friendship isn’t a service. It’s a relationship. And relationships require another person.
This is where Sheridan’s use of Genesis 2 becomes particularly powerful.
The story begins with a remarkably simple observation: “It isn’t good for the man to be alone.” It’s easy to read that as being primarily about marriage, but the idea is much bigger. Human beings aren’t designed for isolation.
And God’s answer to human loneliness isn’t a better communication system, a more sophisticated piece of technology or an infinitely patient companion. God creates another human being.
Someone with their own thoughts, desires, needs, frustrations and freedom.
That matters because one of the greatest gifts another person gives us is their otherness.
A friend isn’t a mirror designed to make us feel good about ourselves. They’re another person whose existence we can’t completely control. They see things differently. They’ve experienced things we haven’t. They introduce us to people we don’t know, challenge our assumptions and occasionally drive us completely mad.
And somehow, we remain friends. That friction isn’t a failure of friendship. It’s part of what makes friendship human.
The temptation of AI companionship is that it offers intimacy without much of the friction.
Research is beginning to suggest why that matters. One 2025 study found that people with smaller social networks were more likely to turn to AI for companionship, while heavier use of AI for companionship, greater self-disclosure and weaker human social support were associated with lower wellbeing. The researchers also concluded that AI companionship didn’t simply substitute for human connection.
That doesn’t mean every conversation with an AI is harmful.
The important distinction is between using AI to help us connect with the world and allowing it to become a substitute for being connected to other people.
That’s a much more important question than whether AI can sound human. Because it can.
The technology is becoming extraordinarily good at producing the language of empathy. It can remember what we’ve told it, respond to our emotions and offer exactly the kind of reassurance we’re looking for.
The machine doesn’t have to feel anything for us to feel something in response and this is where the ethical challenge begins.
The danger isn’t necessarily that we’ll suddenly believe our laptop has a soul. It’s that we might gradually become accustomed to relationships that ask almost nothing of us.
Imagine a friend who always agrees with you, always listens, never interrupts, never asks you to apologise, never has a crisis of their own, never needs anything and never leaves.
An actual friend asks something of us. They need us too.
AI is built around efficiency. Friendship isn’t.
Perhaps that’s one of the most important theological challenges AI presents to us.
We live in a culture increasingly obsessed with optimisation. We want faster answers, better productivity, personalised entertainment, automated administration and frictionless experiences.
Now we can optimise companionship too.
But should we?
What happens if we start treating every relationship as a service that should be available on demand? What happens if we become so accustomed to personalised responses that the ordinary messiness of other people starts to feel like a defect?
We need to be careful about creating a world in which the convenience of artificial companionship makes the inconvenience of human friendship seem like a problem.
Because the inconvenience is part of the gift. Friendship is reciprocal. There’s a difference between being attended to and being needed.
That’s why Sheridan Voysey’s work on friendship is so important at this particular moment. His argument isn’t really anti-technology. It’s an invitation to remember what friendship is before technology starts redefining it.
Friendship Lab asks a wonderfully simple question: Who could you call at 2am when everything has gone wrong?
Its ambition is for every adult to have at least three of these “2am friends”.
I love that definition because it moves friendship away from likes, followers, streaks and constant digital contact.
The deeper question is: Can this relationship teach me how to be a better friend to another human being?
If AI companionship becomes an alternative to human connection, we should be worried.
Perhaps the future of friendship isn’t about finding the perfect companion. Perhaps it’s about becoming the kind of person who’s willing to be one. And maybe the most important question we can ask of artificial intelligence isn’t whether it can become our friend. It’s whether, in a world increasingly filled with machines that can talk to us, we’ll still make the effort to talk to each other.
Because the future of friendship may depend less on what AI learns to say to us than on whether we remember to say: “I’ll be right over.”
No posts

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.