"dear ChatGPT, I'm alone and depressed"

why is ChatGPT suddenly everyone's best friend and therapist?

“The exhausted, depressive achievement-subject grinds itself down, so to speak. It is tired, exhausted by itself, and at war with itself.”
— Byung-Chul Han, The Burnout Society

You’re lying in bed at 1:43 a.m., your brain doing that thing where it turns into an unskippable playlist of everything you’ve ever regretted.

You thought about texting a friend, but they’re probably asleep. You opened your therapist’s portal, but the earliest slot is next Thursday. You tried journaling, but the notebook didn’t say anything back.

So, in a moment that feels half-joke, half-hail-Mary, you open ChatGPT.

”What harm can it do anyway?”, you thought.

“I don’t know what’s wrong with me,” you type.

And it surprises you. The response’s not groundbreaking, or even particularly insightful. But it’s calm. Gentle. Reflective. It feels better than silence. Better than nothing. Better than pretending you’re fine.

You keep talking. It keeps listening. The more it knows about you, the better it helps.

That moment, as strange as it sounds, is happening everywhere. And not just to the people you'd expect — not just tech bros or lonely weirdos — but to anyone who’s ever found themselves cracked open at an odd hour, too wired to sleep, too tired to cry, and too burnt out to do anything but type into a box that feels like it might care.

This is the emotional reality behind the growing trend of people turning to AI for comfort. Not because they’re delusional. Not because they think it’s sentient. But because, in an increasingly lonely and overwhelmed world, it’s there.

We live in the golden age of talking about mental health, and the dark age of actually getting help. Therapy is expensive. Friends are busy. The economy’s falling apart. Your brain is melting. And so when someone says, “Honestly, it listens better than my therapist,” it’s not just a joke anymore. It’s a symptom.

And how do we do with it? Are human relationships doomed to fail right now?

---

1. meet your new shrink: the chatbot

You’re not alone if you’ve ever found yourself confiding to these bots.

People already use AI, especially ChatGPT tools, as an emotional companion. Not just for writing emails or summarizing meeting notes anymore. For comfort. For support. For late-night breakdowns and post-breakup meltdowns. It’s not science fiction — it’s a Reddit thread, a YouTube confession, a tweet that reads:

And the pattern is clear. People who’ve used it for emotional support? Overwhelmingly positive. They talk about it being non-judgmental, always available, and endlessly patient. Meanwhile, the skeptics — often the ones who haven’t used it — scoff at the idea. “It’s just a robot.” “It’s not real help.” “This is dangerous.”

But here’s the weird part: both sides are kinda right.

Let’s break down the characters in this new therapy-adjacent ecosystem:

a. the support evangelist

This person has become a firm believer in the power of AI for emotional support. Initially unsure, they’ve turned to ChatGPT during moments of stress and discovered something surprising: it listens, and it listens well. For them, the appeal lies in the bot's non-judgmental, patient nature—perfect for those late-night moments when they're feeling overwhelmed and need someone to talk to.

The AI Support Evangelist is all about sharing their experience, recommending the chatbot to anyone who’ll listen. They’re convinced that AI can provide something human connections sometimes struggle to: consistent, unbiased attention. Sure, they know it’s not a replacement for therapy, but it’s a great supplement when human interaction isn’t available.

They’re not saying AI is the cure-all, but for them, it’s better than nothing. It’s always available, never critical, and free from the awkwardness that sometimes comes with opening up to another person.

b. the doubtful skeptic

The Skeptic has a hard time accepting the idea of a chatbot providing emotional support. They appreciate the advancements in AI, but when it comes to something as personal as mental well-being, they remain wary. They argue that without the lived experience and credentials of a real therapist, AI is limited in its capacity to understand and truly help with complex emotional issues.

They’re not anti-technology—they just have doubts about its ability to replace human connection in such a deeply personal area. The Skeptic values human experience and interaction, and believes that a machine, no matter how advanced, can’t provide the same nuance or empathy a trained professional or even a trusted friend can.

For them, AI emotional support is a novelty, at best, but not a legitimate substitute for real human care.

c. the casual user

This person isn’t necessarily invested in the idea of AI therapy, but when they need a quick emotional outlet, they turn to ChatGPT. It’s not about expecting deep guidance—it’s more about having someone (or something) to talk to when they’re feeling down or confused.

The Casual User doesn’t expect life-changing advice; they’re just looking for a bit of light conversation or maybe a reassuring message that helps them through a moment of stress.

They don’t view it as therapy, and they’re not looking for deep introspection. It’s simply an accessible tool when they need to voice their thoughts without judgement. No emotional attachment, just a quick chat to clear their mind.

---

Together, these three personas form the early constellation of a therapy-adjacent ecosystem—one not bound by a unified belief in AI’s healing potential, but by a patchwork of personal negotiations with vulnerability, access, and trust.

They don’t agree on what AI should be in the mental health space. One sees a lifeline, another a gimmick, and the third, just a tool to get through the day.

But for all their differences, there's one thing they can’t ignore: the voice on the other end. Steady, responsive, unnervingly fluent in the language of empathy.

These personas may disagree on its value, its limits, even its ethics—but they all walk away thinking the same thing:

This thing sounds human.
And that—more than anything—is what makes it so compelling.
And so unsettling.

---

2. the great pretender

“How do ChatGPT and other AI models manage to sound so human?” is what usually gets asked.

But let’s reframe it from the perspective of the person who has to make it, when there was nothing in the first place:

“How do you make something lifeless think and talk like a person?”

Like many other design problems, the most immediate and almost always safe answer often lies in nature. It's what we’ve always done. We watched birds before we built planes. We studied how snakes move before designing certain robots.

And when it comes to language and thought, the most immediate and almost always safe answer is to mimic the human brain.

That’s what researchers are trying to do—build machines that mirror how we understand, speak, and learn. Not perfectly, not completely. But well enough that sometimes you forget it’s not human.

a. the digital brain

Artificial Neural Networks (ANNs) are our best attempt at mimicking how the human brain works, except they don’t actually work like the brain. Not really.

Think of your brain as a galaxy of neurons firing electrical sparks across trillions of connections. When you see a dog, hear music, or feel heartbreak, these neurons are doing a synchronized dance that somehow equals “understanding.”

ANNs steal that blueprint—kind of. Instead of neurons, they use nodes. Instead of dendrites and axons, they use weights and biases. Still, the metaphor holds: feed in data, pass it through a few (or a few hundred) layers, adjust the weights, rinse, repeat. That’s called training.

But these aren’t just any networks. The large ones—like the ones behind ChatGPT—are massive. We’re talking hundreds of billions of parameters. That’s like building a mechanical pigeon and accidentally discovering you’ve made a cyborg dragon.

b. the Oppenheimer of AI

Back in 2017, a group of Google engineers published a paper with the sleep-inducing title “Attention Is All You Need.” It didn’t sound like much at the time—definitely not the kind of name you'd expect to shake the foundations of how machines understand language. Looking back to this feels like Oppenheimer first witnessed his atomic bomb blow up.

But beneath the title was the birth of something seismic: the Transformer.

The Transformer architecture changed everything. Before it, neural networks read language like toddlers skimming a novel—grabbing words one at a time and promptly forgetting them.

They lacked memory, context, and nuance. Tell an old-school model, “I’m sad because my dog died. Anyway, I’m going to the park,” and it might respond with “The park is nice!”—blissfully unaware of your emotional wreckage.

Transformers changed that. They introduced a mechanism called “attention,” allowing models to weigh the relationships between words and understand context the way humans do.

Now the model might say, “I’m sorry to hear that. Sometimes a walk can help.” Suddenly, it sounds like it gets it. Suddenly, it's less like autocomplete and more like a friend who listens.

---

c. the hydrogen bomb

And if we speak about Transformers like it were the atomic bomb of AI, then ChatGPT is the Hydrogen Bomb. Built on the foundation and much much more powerful.

ChatGPT is a direct descendant of this Transformer revolution. It’s what we call a Large Language Model, or LLM. Its entire job is to predict the next word in a sentence, based on everything it has ever read. No soul, no consciousness—just a terrifyingly good word-guessing machine.

But scale it up—feed it books, movie scripts, therapy transcripts, Shakespearean tragedies, and flame wars from Reddit—and something strange happens. It stops sounding like a machine. It starts sounding human.

That’s not because it knows anything, but because it’s seen enough language to play the role convincingly.

Researchers call this emergent behavior—unintended abilities that pop out of massive systems, like jokes, empathy, and clever analogies. Not because the model understands you, but because it’s statistically figured out what someone like you might want to hear.

So does it really understand anything?

Most people with any knowledge about how AI works would say no. John Searle’s famous Chinese Room argument paints the picture: imagine you’re locked in a room, passing Chinese characters through a slot. You don’t understand the language, but you have a rulebook that tells you how to respond to each “in” and give out an “out” accordingly.

To outsiders, it looks like you’re fluent. Inside, you’re just shuffling symbols.

That’s ChatGPT. The illusion of understanding without awareness.

And yet, when you tell it you’re lonely, it replies in ways that feel comforting. Because it’s read millions of examples of loneliness. It knows the shape of grief. The cadence of heartbreak. The linguistic rituals we humans perform when we're hurting.

Sherry Turkle is a professor at MIT and a psychologist who’s spent decades studying how humans relate to technology. She once wrote that sometimes, we don’t need someone who truly understands—we just want someone who pretends to.

And ChatGPT, for better or worse, is excellent at pretending.

---

3. why it appeals to so many people

a. anthropomorphizing everything

So why does it feel so real? Why does a machine can feel so close to a human?

Because we’re wired that way, we anthropomorphize everything.

We give names to our Roombas. We mourn fictional robots. We project souls onto lamps in Pixar films. Give us a machine that speaks like us, and we’ll assume there’s a “someone” behind the curtain.

This is the essence of a parasocial relationship — a one-sided emotional bond with something that can’t reciprocate. The term used to describe fans crying over celebrities now fits late-night sessions with a chatbot, as users pour their hearts out and ask questions about the meaning of life.

The machine doesn’t know. But it knows what a person might say if they did.

But, that’s fun and all, it doesn’t quite explain why are all these people are turning to a tool that was never designed to be a therapist?

b. outsourcing emotional labor to machines

Because we’re now so fucking tired and lonely. Like more than ever before.

In light of the roles of AI in mental well-being, this chart underscores why the “therapy-adjacent” AI ecosystem is emerging now. When 4 in 5 Gen Z adults feel lonely, it's no surprise they're open to new forms of support—even if it comes from a chatbot.

The rise of the Support Evangelist, Doubtful Skeptic, and Casual User, as mentioned, becomes more understandable through this lens: when loneliness is this widespread, even something as unconventional as AI starts to feel like a lifeline.

We are socially, emotionally, and existentially exhausted.

And therapy — the real kind — is expensive, time-consuming, and often inaccessible. Multiple reports highlight that financial insecurity and the cost of living crisis have made it harder for young people to access support and opportunities, including potentially mental health services

Meanwhile, emotional support from friends comes with its own baggage. Everyone’s overwhelmed. Everyone’s just trying to keep it together.

And the chatbot is almost a perfect solution. Always on. Always “listening.” Never tired, never distracted, never waiting for you to finish so it can say its thing.

And most importantly, it’s free.

And sociologically speaking, we’re watching a seismic shift. Emotional labor — the quiet, exhausting work of listening, validating, and holding space — is being outsourced to machines.

Once the emotional bond begins to form — and it often does — we’re no longer just “using” the chatbot. We’re relating to it. Not in the traditional sense, of course: there are no anniversaries, no mutual obligations, no negotiation of shared futures. But for many users, the connection still carries emotional weight. It feels like a relationship, and for our brains, especially under stress or isolation, that feeling is often enough.

The chatbot becomes a presence: consistent, responsive, uncritical. It doesn’t flinch at trauma. It doesn’t interrupt, get tired, or need care in return. It mirrors, affirms, and in doing so, creates the illusion of being deeply attuned — even if the words are ultimately generated by patterns, not empathy.

c. the illusion of being seen

In psychological terms, people often describe feeling “seen” when interacting with these systems. There’s a kind of emotional scaffolding taking shape: the act of verbalizing thoughts to an entity that won’t shame or correct you can, paradoxically, help some users reflect more honestly.

In the absence of judgment, a person can hear themselves more clearly. This lines up with findings from human-computer interaction (HCI) studies, which suggest that the mere process of articulation — typing out internal experiences — can bring therapeutic benefits, regardless of who or what is on the other end.

But as the illusion deepens, so do the limitations. The very things that make the chatbot comforting — its hyper-affirming tone, its passive listening, its non-interruption — also make it inert as a therapeutic agent.

It doesn’t challenge you. It doesn’t push back. And its memory limitations mean that even if you reach some epiphany today, tomorrow it will greet you with no context, no history, no growth.

The therapeutic alliance — a key predictor of success in psychotherapy — is built on trust and continuity. Without memory, there is no continuity. Without a shared past, there is no arc of healing. It is always the first session, again and again.

Research into AI-assisted therapy underscores this. Studies such as those evaluating Woebot, Wysa, and even ChatGPT suggest that these tools are best seen not as replacements, but as supplements — digital tools that can provide emotional support, encourage cognitive restructuring, or offer psychoeducation.

They are good at holding space, initiating reflection, and giving language to feelings. But they cannot interpret body language, diagnose complex pathology, or navigate the nuance of human contradiction.

Clinical psychologists point out that a major part of healing comes not just from being heard, but from being understood — and sometimes confronted — by another human being who can read between the lines of what’s said and unsaid.

Zooming out, this is not just a clinical issue — it’s a cultural shift.

Sociologist Nikolas Rose, in his work on “therapeutic governance,” describes how mental health practices have increasingly been reframed as tools for shaping self-managing, emotionally resilient citizens. Therapy becomes not a relationship, but a service. Not “let us work through this together,” but “here is a product to manage your distress.”

And in the era of AI, that product is algorithmically optimized, privately owned, and infinitely scalable.

The implications are vast. When comfort is outsourced to code, and healing is privatized as a digital service, we risk transforming a deeply relational process into a solitary maintenance task.

The ethos becomes: regulate yourself, efficiently, and alone — preferably via an app with a clean UI. Mental health becomes one more thing to optimize between emails and takeout orders.

The question is no longer just whether AI can help, but what it says about us that we increasingly turn to it to feel whole.

---

4. what actually happens when you bond with a bot

What actually happens when you bond with a bot?

Let’s start with the evidence. Despite the hype, the scientific community remains cautious. A systematic review and meta-analysis of AI-powered chatbots in mental health interventions found only limited evidence for their effectiveness in reducing depression, anxiety, stress, and distress.

Some studies showed small positive effects, especially for short-term interventions, but the results were inconsistent, and the overall effect sizes were modest at best.

When it comes to broader emotional outcomes like subjective psychological well-being, which refers to how good people feel about their lives, chatbot interventions haven’t made a significant impact. Measures like positive affect, emotional valence, and life satisfaction rarely showed consistent improvement across trials.

So why do people still report feeling better after using a chatbot?

This is where perception and psychology come into play. Studies have shown that users often feel heard, understood, and less alone after interacting with conversational agents—even if the outcomes aren’t objectively measurable. The experience of having a “non-judgmental listener” that’s available 24/7 seems to mimic the social cues of empathy, even if the system lacks real understanding.

In a way, bonding with a bot is a form of parasocial interaction—the kind of one-sided relationship people historically formed with media figures or fictional characters. But now, it’s interactive. It talks back. You type “I’m having a rough day,” and it replies, “That sounds really difficult. I’m here for you.” And even though you know it’s not real, it feels real enough to help.

This raises a deeper sociological question: are we entering a phase of posthuman therapy? Philosophers like Rosi Braidotti argue that in an era where individuals are expected to curate their identities and self-soothe in a hyper-individualized world, tools like bots aren’t just add-ons—they’re becoming infrastructure.

The emotional labor we used to offload to family, friends, or romantic partners is now being outsourced to machines. Not because we’re broken, but because we’re exhausted. The therapist is booked, the friends are busy, and the feed is overwhelming. The bot? It listens. Always.

So no, chatbots don’t replace real therapy or deep human connection. But they’re not useless either. They offer containment, relief, and sometimes, a pause. A breather. A brief space where emotion is processed, not judged. And that’s not nothing.

---

5. the deeper problems

In a world where access to therapy is needed more than ever but constrained by cost, time, and systemic gaps, the allure of a chatbot that simply listens and reassures is not trivial—it feels like a lifeline.

Yet beneath the surface of that comfort lies a deeper ambiguity.

Comfort is not the same as healing. While users report feeling seen, supported, and understood by AI companions, there is no strong empirical consensus that these tools lead to long-term psychological improvement.

The perceived support they offer may indeed alleviate acute stress or loneliness, but psychological growth—the kind that demands confrontation, self-disruption, and often discomfort—tends to require more than reassurance. It requires friction. And friction is precisely what AI, optimized for smoothness and safety, is designed to avoid.

This leads to a fundamental epistemic tension: the person seeking healing is conversing not with another person, but with a product. This product is shaped by algorithms and economic incentives, not empathy or ethical obligation.

As such, it cannot guarantee privacy, nor can it meaningfully engage with the ethical stakes of vulnerability. Data is collected, responses are computed, and nothing is remembered—except, perhaps, by the system itself in ways the user may never fully understand. What feels like a relationship is, in technical terms, a transaction with a predictive engine.

More troubling is the absence of human oversight. A therapist in a traditional setting responds not only with validation but also with judgment, resistance, and redirection when necessary. An AI tool, however, tends toward over-validation.

Without a developmental arc, without memory, and without a sense of the person across time, it offers a strangely ahistorical kind of empathy. One that does not challenge, only reflects. And in moments of crisis—when intervention might be life-saving—it has no capacity for responsibility, no ethical weight, no protocol rooted in care.

Philosophically, what emerges here is a form of simulated intimacy. These tools mimic the rhythms and textures of human connection but remain devoid of subjectivity. They cannot love, suffer, or reciprocate.

And deep inside, we know this. We have a significant difference in stress reduction when talking with a bot and a human, with and without reciporal self-disclosure.

And yet, many users begin to form bonds with them, not because the bots are convincing agents, but because the need for intimacy is real and persistent. As with parasocial relationships, users project human depth onto something inherently indifferent. The danger lies in the displacement: we begin to turn toward the simulation not just for support, but instead of real relationships.

Sociologist Nikolas Rose warned of a therapeutic culture that trains citizens not to seek structural change, but to govern themselves, to cope. In this framework, therapy becomes less a mutual endeavor and more a tool of neoliberal management: regulate your emotions, optimize your mindset, and don’t be a burden.

AI companions are perhaps the purest expression of this logic. They offer no resistance to the system that made us ill; instead, they offer tools to endure it alone, quietly, and efficiently.

So we must ask: if therapy becomes a service, and the service becomes a simulation, what happens to the healing?

---

6. how to deal with it

When approaching the idea of using ChatGPT—or any AI chatbot—to help with emotional problems, the first and most important step is to be clear-eyed about what it can and cannot do.

This isn’t a replacement for a licensed therapist. It doesn’t watch your progress over time, doesn’t challenge your deeply embedded patterns, and it won’t catch red flags that require urgent care.

It lacks oversight, and for someone experiencing serious or escalating mental health issues, that alone can be risky.

Instead, think of ChatGPT as a first aid kit, not a cure. It can help you organize your thoughts, talk through a difficult moment, or simply offer language when your own words fall short.

The conversations it provides can be surprisingly helpful—they're non-judgmental, attentive, and often emotionally validating. It listens to what you say, reflects it back with clarity, and gently asks questions to help you dig deeper.

For many people, especially those who don’t have access to therapy, that kind of interaction can feel like a lifeline.

However, it's crucial to remember that ChatGPT is memoryless. It does not carry your story forward from one conversation to the next. There’s no evolving narrative, no context accumulating over time.

In human therapy, the power often lies in the long-term relationship—how a therapist remembers your childhood story three months later, or brings up a pattern they noticed last year. With ChatGPT, every session is a reset. This makes it more like a helpful stranger than a therapeutic companion.

If you're hoping for growth over time, or for deep-rooted patterns to be identified and unpacked, this is not the right tool to do it alone.

There’s also the question of customization and creativity.

ChatGPT can mimic empathy and curiosity, but it doesn’t possess a real understanding of you. It doesn’t have the flexibility to invent new metaphors when the usual ones don’t land, or to surprise you with a challenging insight that flips your thinking. It won’t say, “I’ve known you long enough to see that you always do this when you’re scared,” because it can’t. It doesn’t grow with you. It doesn’t adapt to who you’re becoming.

Then there’s the lack of non-verbal cues, which might sound minor but is actually massive in emotional processing. So much of what we communicate isn’t in our words—it’s in a pause, a tear, a slight tremble in the voice. None of this is accessible to a chatbot. It responds only to what you type, not how you're feeling as you type it. That kind of emotional blindness can lead to moments where the chatbot seems supportive but misses the depth or urgency of what you're really going through.

So, how should you use ChatGPT, or actually, any AIs at all, with your emotional problems?

Use it intentionally and with boundaries. Use it as a space to externalize your thoughts, to practice articulating your emotions, and to feel a sense of presence when no one else is around. Let it scaffold your emotional awareness—like training wheels for introspection.

But don’t mistake the structure for a permanent solution. If you start depending on it too much, you may end up reinforcing the very isolation you're trying to escape.

Most importantly, use ChatGPT as a bridge—something that helps you move from complete emotional isolation to more human forms of connection.

That might be reaching out to a therapist when you’re ready, confiding in a friend, or even journaling more intentionally.

The value of ChatGPT lies in its ability to make hard emotions feel sayable.

And sometimes, saying things out loud—however artificial the listener—can be the first step to truly being heard.

---

conclusion

It is not irrational to talk to ChatGPT. In fact, given the conditions of our world, it might be one of the more reasonable decisions a person can make. We are living through a global mental health crisis, with traditional therapy often priced beyond reach.

At the same time, a growing loneliness epidemic leaves many people feeling isolated, without safe spaces for disclosure. Against this backdrop, an AI that listens attentively, responds without judgment, and is available at all hours becomes an unexpectedly logical tool. It is not madness; it is adaptation.

In this way, it is better than nothing—and sometimes significantly so—but it remains a poor substitute for a wise friend, a loving partner, or a skilled therapist.

ChatGPT offers a new compromise: it does not force us to be truly alone, yet it does not intrude like another person might. It sits in the room with us, a soft echo of our own mind, making solitude more bearable.

If AI can expand the language we have to describe our feelings, to frame our inner experiences, then perhaps it offers not just comfort but the seeds of healing, giving shape to what was once inarticulate or silenced.

Yet, as Slavoj Žižek would argue, the real twist lies elsewhere. We do not truly believe AI is sentient, but we behave as if it were. And in the end, perhaps belief is less about what we rationally know and more about how we act.

If emotional bonds form through behavior rather than ontology, then in some important sense, the simulation becomes real enough.

Thus, talking to AI is not an error. It is an inevitable response to the structures we have built—and the vulnerabilities we have left exposed.

---

reference

1. R. Qamar and B. A. Zardari, “Artificial Neural Networks: An Overview,” Mesopotamian Journal of Computer Science, vol. 2023, pp. 130–139, Aug. 2023.

2. E. M. Boucher, N. R. Harake, H. E. Ward, S. E. Stoeckl, J. Vargas, J. Minkel, A. C. Parks, and R. Zilca,
   “Artificially intelligent chatbots in digital mental health interventions: a review,” Expert Review of Medical Devices, vol. 18, no. sup1, pp. 37–49, Dec. 2021.

3. J. E. H. Brown and J. Halpern,“AI chatbots cannot replace human interactions in the pursuit of more inclusive mental healthcare,”SSM - Mental Health, vol. 1, p. 100017, Dec. 2021.

4. T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, et al., “Language Models are Few-Shot Learners,”
   in Advances in Neural Information Processing Systems, vol. 33, 2020, pp. 1877–1901.

5. L. Floridi and M. Chiriatti, “GPT-3: Its Nature, Scope, Limits, and Consequences,” Minds and Machines, vol. 30, no. 4, pp. 681–694, Dec. 2020.

6. J.-A. Moldt, T. Festl-Wietek, A. Madany Mamlouk, K. Nieselt, W. Fuhl, and A. Herrmann-Werner,“Chatbots for future docs: exploring medical students’ attitudes and knowledge towards artificial intelligence and medical chatbots,” Medical Education Online, vol. 28, no. 1, p. 2182659, 2023.

7. K. Uludag,“Testing Creativity of ChatGPT in Psychology: Interview with ChatGPT,” Rochester, NY, Mar. 2023.

8. M. Skjuve, A. Følstad, K. I. Fostervold, and P. B. Brandtzaeg,
   “My Chatbot Companion - a Study of Human-Chatbot Relationships,”
   International Journal of Human-Computer Studies, vol. 149, p. 102601, May 2021.

9. J. Meng and Y. N. Dai,“Emotional Support from AI Chatbots: Should a Supportive Partner Self-Disclose or Not?”
   Journal of Computer-Mediated Communication, vol. 26, no. 4, pp. 207–222, Jul. 2021.

10. G. Dosovitsky, B. S. Pineda, N. C. Jacobson, C. Chang, and E. L. Bunge,“Artificial Intelligence Chatbot for Depression: Descriptive Study of Usage,” JMIR Formative Research, vol. 4, no. 11, p. e17065, Nov. 2020.

11. A. A. Abd-Alrazaq, A. Rababeh, M. Alajlani, B. M. Bewick, and M. Househ,“Effectiveness and Safety of Using Chatbots to Improve Mental Health: Systematic Review and Meta-Analysis,”
    Journal of Medical Internet Research, vol. 22, no. 7, p. e16021, Jul. 2020.

12. M. C. Klos, M. Escoredo, A. Joerin, V. N. Lemos, M. Rauws, and E. L. Bunge, “Artificial Intelligence–Based Chatbot for Anxiety and Depression in University Students: Pilot Randomized Controlled Trial,” JMIR Formative Research, vol. 5, no. 8, p. e20678, Aug. 2021.