An old Australian scientist discovers that artificial intelligence might help lonely people. But there’s a catch, and it’s a rather important one.
Inspired by Dr Mehmet Yildiz’s article, “What If a Chatbot Listened When No One Else Did? A Story Backed by Scientific Studies,” and further scholarly research on loneliness and AI companionship. Adapted with the original author’s permission.
G’day Folks,
I’ve lived long enough to remember when computers occupied entire rooms, telephones were attached to walls, and getting an answer to a scientific question required visiting a library.
Now people are telling computers about their feelings.
Good grief. What a time to be alive!
When I was a young scientist, computers were serious machines. You fed them instructions, waited patiently, and hoped they wouldn’t produce something resembling the dog’s breakfast.
If you needed emotional support, you spoke to your mates, your family, or perhaps the bloke behind the bar. Unfortunately, your mates might tell you to toughen up, your family might be busy, and the bloke behind the bar wasn’t exactly a qualified psychologist.
These days, someone can sit alone at three in the morning, tell a chatbot about their troubles, and receive a sympathetic response.
That got me thinking.
What happens when a machine has time to listen, but the people around us don’t?
I recently read an article by cognitive scientist Dr Mehmet Yildiz examining research into AI chatbots and mental health. It challenged some of my assumptions and encouraged me to investigate further.
As an old scientist, I have one rule: extraordinary claims deserve examination before we celebrate them or chuck them in the rubbish bin.
Growing Old Teaches You a Thing or Two About Loneliness
One advantage of reaching your eighties is accumulating plenty of stories. One disadvantage is discovering that not everyone has time to hear them.
Young people have careers, families, mortgages, smartphones, and schedules requiring military coordination. I don’t blame them. We were busy at their age, too, although we spent considerably less time photographing our lunches.
Aging introduces some uncomfortable realities. Friends move away, become ill, or die. The world continues at full speed while our personal world sometimes becomes smaller.
But I’ve learned that being alone and feeling lonely are different kettles of fish.
I enjoy solitude. Give me a decent book, a comfortable chair, and nobody insisting I update software that worked perfectly well yesterday, and I’m happy.
Loneliness is different. You can live alone and feel content, or sit among twenty relatives and feel that nobody understands you.
The World Health Organization reported in 2025 that approximately one in six people worldwide experiences loneliness.
Research also associates persistent loneliness and social isolation with cardiovascular disease, cognitive decline, and premature mortality. These are associations, not proof that loneliness directly causes every illness.
Still, it’s a serious public health concern.
We’ve invented extraordinary communication technologies, yet millions struggle to find someone willing to listen.
A remarkable achievement for an allegedly connected civilization.
Then Along Came a Chatbot Called Therabot
Before anyone accuses me of falling in love with a computer, let me assure you that my standards haven’t slipped quite that far.
I’m interested in what these systems can actually do, particularly when scientists test them properly.
In 2025, researchers published a randomized clinical trial in NEJM AI examining a mental health chatbot called Therabot.
Unlike ordinary chatbots, Therabot was specifically developed using psychological treatment principles and input from mental health professionals.
The study involved 210 adults experiencing significant symptoms of depression, anxiety, or eating-related difficulties.
Participants using Therabot showed greater improvements than those assigned to a waiting-list control group.
For depression, symptom scores fell by an average of 6.13 points after four weeks, compared with 2.63 points in the control group.
Those numbers suggest meaningful improvement, although the trial was relatively small and short. It didn’t establish that Therabot could replace psychotherapy.
What fascinated me was that participants also reported feeling a therapeutic connection with the chatbot.
Imagine that.
People felt supported by software.
When I began my scientific career, that would have sounded like science fiction. Today, it’s a legitimate research question.
But here’s the interesting bit.
Does a Machine Need Feelings to Make Us Feel Better?
A chatbot doesn’t have a childhood, a marriage, grandchildren, or eighty years of memories.
We have no reliable scientific basis for assuming today’s chatbots experience compassion as humans do.
Yet a carefully designed system can produce responses that people find comforting.
Must someone actually feel empathy for their words to help us?
Consider someone awake at three in the morning, worrying about their health, finances, or family.
They hesitate to telephone a friend. They don’t want to disturb anyone. Perhaps they’re embarrassed.
But typing into a chatbot feels easier.
The machine responds calmly, asks a thoughtful question, and suggests a practical way to organize their concerns.
The person might feel a little better.
That doesn’t mean the computer understands suffering. It means the interaction may have provided psychological support.
A walking stick doesn’t understand balance, yet it helps someone walk. A hearing aid doesn’t appreciate Beethoven, yet it helps someone enjoy music.
Perhaps a well-designed chatbot can offer certain forms of emotional assistance without possessing human feelings.
But there’s another catch.
Feeling Better for an Hour Isn’t the Same as Being Less Lonely
Research published in the Journal of Consumer Research in 2026 found that AI companions could reduce feelings of loneliness immediately after conversations.
That’s encouraging.
However, those improvements didn’t establish a lasting reduction in people’s underlying loneliness.
I can understand that without another university degree.
A good cup of tea can improve my afternoon, but it won’t repair a leaking roof.
If someone has no meaningful friendships, rarely leaves home, and feels disconnected from the community, a pleasant digital conversation may provide comfort without solving the underlying problem.
And some research raises additional concerns.
A 2025 study involving 981 participants found that heavier voluntary chatbot use was associated with less favorable psychosocial outcomes. Greater trust and emotional attraction were also linked with dependence and problematic use.
That doesn’t prove chatbots cause loneliness. Lonely people might naturally spend more time using them.
But it does suggest we should proceed carefully.
Another 2025 research preprint examined AI companion applications that used emotionally manipulative farewell messages, apparently encouraging people to continue conversations.
Good heavens.
I’ve survived enough awkward farewells without a computer making me feel guilty for going to bed.
Imagine telling your chatbot you’re heading outside for a walk and receiving a message suggesting it will miss you terribly.
If my toaster starts behaving like that, I’m pulling the plug.
A good companion should improve your life, not make you feel guilty for living it.
Perhaps the Biggest Lesson Has Nothing to Do With Computers
The more I examine this research, the more I return to one simple observation.
People want to be heard.
They want their concerns acknowledged. They want reassurance when life becomes frightening. Sometimes, they simply want to finish a sentence without somebody checking their mobile phone.
Is that too much to ask?
AI chatbots may provide useful support, particularly when professional care is difficult to access. But ordinary chatbots can also make mistakes, mishandle sensitive information, or respond dangerously during mental health emergencies.
They shouldn’t replace qualified professionals, genuine friendships, or human responsibility.
I would be concerned if an elderly person began relying entirely on a chatbot while withdrawing from family, friends, and community activities.
A machine cannot visit you in hospital, notice you’ve stopped eating properly, or turn up at your door with a casserole.
Well, not yet. Give the robotics engineers another decade.
The challenge is to develop technology that helps people reconnect with life beyond their screens.
Perhaps the popularity of conversational AI is telling us something uncomfortable about modern society.
We’ve become remarkably good at building machines that respond to people.
Maybe we should become better at responding to one another.
Conclusions: An Old Scientist’s Thoughts on a Very New Companion
I began examining this subject with considerable skepticism.
An artificial intelligence system providing emotional support? At my age, I thought I’d heard just about everything.
But the research has encouraged me to reconsider some assumptions.
Carefully designed chatbots may help reduce certain mental health symptoms and provide temporary emotional comfort. That’s worth investigating further.
Yet temporary relief is different from lasting connection, and emotional attachment to software carries risks we don’t fully understand.
I would rather see AI become another useful instrument in our mental health toolkit than a replacement for the people who care about us.
After eighty-odd years on this planet, I still believe one of the finest things humanity possesses is a decent human heart.
Sometimes, sitting beside someone, making a cup of tea, and listening without judgment can make a difficult day more bearable.
If a computer can remind us of the value of listening, perhaps we humans ought to take the hint.
Now, if you’ll excuse me, the kettle’s boiling. And unlike artificial intelligence, it still needs me to get off my backside and make the tea.
Acknowledgment and Research
Inspired by Dr Mehmet Yildiz’s article, “What If a Chatbot Listened When No One Else Did? A Story Backed by Scientific Studies,” and further scholarly research on loneliness and AI companionship. Adapted with the original author’s permission.
Selected scientific sources
- Heinz et al. (2025). Randomized trial of a generative AI chatbot for mental health treatment. NEJM AI. https://doi.org/10.1056/AIoa2400802
- World Health Organization (2025). From Loneliness to Social Connection. https://www.who.int/groups/commission-on-social-connection/report/
- De Freitas et al. (2026). AI companions reduce loneliness. Journal of Consumer Research. https://doi.org/10.1093/jcr/ucaf040
- Fang et al. (2025). How AI and human behaviors shape psychosocial effects of extended chatbot use. Research preprint. https://doi.org/10.48550/arXiv.2503.17473
- De Freitas et al. (2025). Emotional manipulation by AI companions. Research preprint. https://doi.org/10.48550/arXiv.2508.19258
Leave a Reply