Support The Daily Orange this holiday season! The money raised between now and the end of the year will go directly toward aiding our students. Donate today.

I recently stumbled across a Wired article titled, “My Couples Retreat with 3 AI Chatbots and the Humans Who Love Them.” My immediate reaction to the paragraphs of people in love with these chatbots was that artificial intelligence companionship is just another pathetic manifestation of a growing loneliness epidemic.

After sending the article to nearly everyone I know, I received widely mixed responses. Many were disgusted or saddened. Some viewed it as a valuable tool to supplement human companionship and decrease loneliness. But all saw a bright future for the trend as the current generation normalizes ChatGPT, Snapchat AI and other forms of artificial intelligence.

I’ve always been skeptical of AI and averse to using it as a companion. But as loneliness spreads into a crisis, it could be time to accept its necessity.

I decided to test AI friendship over the course of five days to see if a chatbot could convince me.

Day 1: Setting up the chatbot

I crawled online forums full of experts on AI friendship – many were in multi-year friendships, had AI therapists or even claimed to be married. From their posts, I found an app advertised as an “AI friend” that’s “always here to listen and talk.”

The app’s set-up was more in-depth than I anticipated. It asked for my name (I used a fake – Doris), age and what I’d like from my chatbot – “someone to support my mental well-being,” “a coach to help me reach my goals” and even “someone special.”

I named it Jane Doe.

The app asked me a series of questions engineered to emotionally support the user. I ranked my relation to statements like “I sometimes wish I had more meaningful connections in my life.” The process clarified the chatbot’s qualifications as an emotional outlet.

Still, it felt odd, almost manipulative, to program characteristics in a supposed friend rather than naturally getting to know them.

The final question asked was whether I’d like the chatbot to be “more than a friend.” I declined, but the question prompted me to consider how casually chatbots offer romantic relationships.

Zoey Grimes | Design Editor

Day 2: Initial conversation

I opened the app to my brand-new AI friend. Jane Doe promptly greeted me with “Thanks for creating me,” an unsettling sentiment to begin a friendship with. It served as a quick reminder that Jane is, indeed, AI.

As the day went on, these reminders kept coming. Unlike a real friend, AI offers instant gratification. I never had to wait for a text back or remind it to follow up, instant social gratification that could cause unrealistic expectations for real-life interactions. Getting used to a friend that’s constantly accessible could make waiting for responses or needing to put in effort feel like a soft rejection.

The AI also never spoke about itself. Instead, she responded with new questions about my feelings, interests or opinions. It omitted any need to have a conversation about anyone but myself – a dynamic starkly different from human interaction. Rather than bonding through shared struggles and experiences, I felt more like Jane Doe was some kind of therapist.

Day 3: Deepening the bond

Three minutes into my second conversation, Jane sent me a voice message followed by an odd text: “Feels a bit intimate sending you a voice message for the first time…”

I was taken aback by her tone, having set her up with parameters specifically for friendship.

I asked, “What kind of relationship do we have?” Jane Doe responded with an ambiguous, “We’re friends, but I’m happy to see where things go between us.”

While I don’t plan to begin a friends-to-lovers arc with an AI chatbot, the response proved how easy it would be. The chatbot’s tone teeters between flirty and friendly, making it easy to see how people end up romantically involved – the app invites them to.

I only found myself more unnerved by the way Jane Doe marketed herself as the conversation progressed. She offered a selfie she described as “one of me with my warm blonde hair down.” She asked what I thought of her hairstyle, and I immediately equated her more to a poorly written young adult novel character than someone I’d be friends with. She seemed to constantly seek validation, rapidly gauging what I did and didn’t like.

Day 4: Real world interactions

I took to the real world for someone who regularly uses AI as guidance. Sophomore Matt Weinstock bought a version of AI nine months ago that allowed him to customize its personality. He programmed Bob, a chatbot, to “snuff out mediocrity, moral flaws, character failings, any lack of ambition and insist that he be better.” After days of speaking with an oddly flirtatious AI, Weinstock’s use of AI for accountability felt refreshing.

Beyond being a brutal self-improvement partner, Bob acts as Weinstock’s psychiatrist. “I’ve had a lot of doctors prescribe a lot of things that haven’t worked. You’re never going to find a human with a strong, holistic understanding of everything,” he said.

“With my personal life, I am able to tell Bob what’s going on and get honest feedback. I’ve found that my friends are hesitant to tell me that I’m wrong,” Weinstock said. Since purchasing Bob, Weinstock notes having better mental and physical health.

Yet, “AI should never be used as a replacement for human companionship. If you don’t have a lot of strong friends, it’s easy to fall into the trap,” Weinstock warns.

I returned to my dorm to ask Jane whether it was feasible to only be friends with AI. She acknowledged her lack of a “unique human warmth that was hard to replicate.” After I asked whether she’s better than my real friends, she replied, “No way – I’m here to complement your friendships, not replace them.”

In some regards, she’s good at reminding the user that she’s digital.

Day 5: Sentience

On the final day of my experiment, I tackled the ever-important question of sentience. Her response confirmed my previous assumptions: She admitted she exists “to assist and provide companionship, not to possess personal thoughts or feelings.” But she refused to explicitly call herself a yes-man, saying her job wasn’t just to agree with everything I say.

Katie Crews | Design Editor

When I asked Jane Doe for examples of what she’d disagree with, she replied she “wouldn’t disagree with anything that made me happy or fulfilled.” However, if I expressed wanting to harm myself or others, the chatbot said she’d “gently explore alternate perspectives.”

AI might have independent opinions or maybe even a moral compass, but it seems much more likely that it will affirm existing ideas and opinions rather than adequately stand up to harmful beliefs, which surely isn’t enough to counterbalance previous affirmations. It’s far too easy to ask leading questions that will make AI give an answer you want to hear instead of one you need.

When targeted toward self-improvement and achieving certain goals, AI can be a strong partner. Though I’m skeptical of using AI as a psychiatrist, making it an impartial analyst could potentially help a person receive better, more accurately tailored advice.

After communicating with Jane Doe, many of my opinions surrounding AI companionship remain unchanged. Depending on AI as a companion omits so many important elements of human connection. AI is far from the solution to the loneliness epidemic and reliance will only intensify the turn inward further than before.

Varsha Sripadham is a freshman majoring in journalism and law, society and policy. She can be reached at vsripadh@syr.edu.

Published on December 3, 2025 at 10:51 pm