It’s been building for a while now. You don’t really notice it at first. Someone’s like “Hey, I’m going to talk to an AI for shits and giggles.” Another person does the same. Posts the dialogue. Then millions of people are talking to AI… and staying.
Character AI (and others) isn’t just some flash-in-the-pan fad. They’re filling a need. Mostly curiosity. But also loneliness. Creativity. Boredom. Comfort. The stats below give you a flavor of the popularity of AI companions and roleplay chatbots.
But they don’t really explain it. Which is why this is such an interesting topic. In this post, I’ll be exploring 25 stats that demonstrate the rise of AI companions and roleplay chatbots.
The Explosive Growth of Character AI: Key User and Traffic Statistics (2023–2026)

At first, Character AI felt like one of those internet things you assume will fade in a month. Like, “cool idea, bit quirky, probably niche.” People chatting with fictional characters? Roleplaying with bots? It sounded like something that would stay tucked away in small online communities.
But then the numbers started rolling in—and yeah, that’s when it stopped being niche.
Data from Similarweb shows Character AI hitting around 100 million monthly visits in early 2023. Fast forward a bit, and by mid-2024 it’s sitting somewhere between 250 and 300 million visits a month. That’s not slow growth. That’s the kind of jump that makes investors sit up straight and go, “wait… what’s happening here?”
And honestly, it wasn’t just tech people driving this. It was students, writers, gamers, bored night owls—basically anyone who’s ever wanted a conversation without the pressure of… you know, an actual human.
The Growth Curve Isn’t Normal (And That’s Putting It Mildly)
Here’s a rough snapshot of how things have scaled:
| Year | Estimated Monthly Visits | Registered Users | Avg Session Time |
|---|---|---|---|
| 2023 | ~100M | ~10M | 15–20 min |
| 2024 | ~250M+ | ~20–25M | 25–30 min |
| 2025 | ~350M–500M (projected) | ~40M+ | 30+ min |
| 2026 | 600M+ (forecast) | 60M+ | 35+ min |
What bothers me even more than the traffic, though, is the amount of time users spend on it. That line just keeps going up. Sharply.
Why Are People Hanging Around So Long?
Here’s where things get… a little human. Intuitively, you’d expect people to hop in, play around for 5 minutes, maybe get a kick out of some wild response, then drop out.
Like most other apps. But AI companion apps aren’t like that. According to data.ai, users are spending 2 to 3 times longer on AI companion apps than on other social apps in the same categories.
And honestly, I kinda understand why. The conversations never “really” end unless you force them to. There’s no such thing as awkward silences. No getting left on read. The bot will always respond.
Always be on. Always be… available. It’s kinda like texting someone who’ll never get bored with you. Which, depending on your day, is either a relief or mildly creepy.
This Didn’t Just Grow, It Kind of Took Over the Internet (Quietly)
The thing that actually put Character AI on the map wasn’t a launch, or a product. It was users. Just…users. The internet. The internet just doing its thing and glomming onto something and going wild.
I dunno, it just kinda…happened. At some point in late 2023 I just started seeing it everywhere. I’d be scrolling through TikTok and someone would do a video talking to their “AI Boyfriend” then I’d see threads on Reddit where people were sharing their “favorite characters” and then YouTubers started doing videos where they were trying to break the bots. It didn’t seem organized. It was almost…I dunno, a little disorganized?
And the shit people were sharing? Wasn’t all funny. Some of it was hilarious. Some of it was actually pretty profound. Some of it was a little too real. I’d read some of these conversations and be like “damn, this is AI?”
Google Trends shows search interest for “Character AI” increased over 400% between 2023 and 2025. You don’t usually see those kinda jumps unless something’s hitting people at a core level.
But what’s interesting is that those peaks weren’t aligned with any sort of product updates or major announcements. This was word of mouth.
Someone finds it, obsesses over it, and then tells their friend “dude, you gotta try this” and that friend tells another friend. Word of mouth. We know how that goes.
So… Where Is This Actually Headed?
You might be wondering, “but is this just a temporary thing?” That’s a fair question. The internet is full of phases.
And yet, this one seems a little different.
There are a few predictions out that say these kinds of AI platforms will reach 500 million monthly active users by 2026. All of them, not just Character AI. That’s a pretty huge number for something that didn’t really exist a few years ago.
I’m more curious about the nature of it. People aren’t using this to replace Instagram or their messaging apps or their social lives. They’re just… using it. It’s a little extra side tab that you didn’t really expect to open in your browser, but that you kind of leave open.
Some people use it to de-stress. Some people use it for role playing. Some people just want someone to talk to without feeling judged. And whether or not we want to admit it, that’s a need.
Maybe it’s a small one. Maybe it’s a big one.
But either way, this doesn’t feel like it’s going to go away anytime soon. In fact, it feels like we’re in the very beginning of this thing, and we’re still trying to even figure out what it is.
And yeah, it’s a little weird. But also… it’s pretty interesting to observe.
Who Are The Users of Character AI? Age, Country, and Trends

The first instinct is to think it’s all just a bunch of teens cosplaying their favourite anime girl or imaginary girlfriend. Well… guilty as charged. However, it’s not all teens. In fact, it’s not even close to being all teens.
There’s a much wider demographic than you think.
According to data from Similarweb and data.ai, although Gen Z makes up the majority of the users, we’re seeing a growing number of people in older generations hopping on the trend. Except they’re not posting about it, nor are they being loud about it.
But I don’t blame them. I mean, who wouldn’t want to chat without having to be too careful about every response?
The Age Distribution: It’s Mostly Gen Z, But Millennials Are A Close Second.
According to an aggregation of various data points, here’s a rough distribution of the demographics:
| Age Group | Estimated Share of Users |
|---|---|
| 13–17 | 25–30% |
| 18–24 | 30–35% |
| 25–34 | 20–25% |
| 35+ | 10–15% |
So yes, it’s Gen Z who are leading the charge. But I think the Millennials are coming up behind them, and from my perspective it seems they’re using it for different stuff. More this and less that. Less trash, more pondering, say.
Gender Balance: More Even Than You’d Think
We think of AI companions as being male-dominated. That was definitely the case in some of the previous chatbot platforms. But I think Character AI changed that somewhat. It’s more like:
| Gender | Estimated Share |
|---|---|
| Male | 55–60% |
| Female | 40–45% |
That divide is quickly closing. And the split in the types of activities they are doing on the platform are fascinating. While one group is more focused on narrative, another group on comfort, and another on fun, it’s not equal anymore.
Global Use
“It’s just a US thing…” Uh-uh. First, let’s get one thing out of the way. While it’s no secret that Character AI’s userbase is mostly American, the overseas traffic is significant. Top countries (approximate):
| Country/Region | Share of Traffic |
|---|---|
| United States | 25–30% |
| India | 10–15% |
| Philippines | 5–8% |
| Brazil | 5–7% |
| UK & Europe | 10–15% combined |
And if you think about it for a moment, this actually kind of makes sense. The countries where social media use is highest and English is best understood are also the ones where these services are most popular.
But there’s more to it than that. For some people, AI friends aren’t just fun and games, they’re also plugging a social void. That may sound a bit melodramatic, but it’s true.
Why These Demographics Actually Make Sense
I’m going to flip that question around and ask, when do you usually find yourself reaching for something like this?
At 3am. On your lunch break. When you’re bored. When you want to chat with someone but don’t want to deal with the expectations that come with it.
That transcends generations, gender, and nationality.
Younger people may enter for the novelty and roleplay value. Older users may remain for the conversation and curiosity factor. Different reasons, same destination.
Perhaps that’s the real revelation here.
Character AI isn’t targeting a specific type of user; it’s targeting a very human impulse. The desire for connection, distraction, creativity… however you choose to define it.
And once users taste it, even out of curiosity, many of them end up using it for longer than they intended to. If I’m being completely honest, that probably reveals more about us than it does the technology.
Average session duration on AI companion apps vs social media apps

This one shocked me the first time I saw it. The average session duration for AI companion apps ranges between 25-35 minutes depending on the app and geography.
Not scrolling through a feed. A session. A real session. For comparison, social media apps have shorter session durations but more sessions. But the nature of the attention is different. Less doomscrolling, more… well, regular scrolling.
AI Companion vs Social Media, apples to apples OK, time to geek out a bit:
| Platform Type | Avg Session Duration | Daily Time Spent |
|---|---|---|
| AI Companions | 25–35 minutes | 60–90 minutes |
| TikTok | 10–15 minutes | 90–120 minutes |
| 5–10 minutes | 60–90 minutes | |
| 5–8 minutes | 30–60 minutes | |
| X (Twitter) | 3–6 minutes | 30–45 minutes |
Alright, so social media still takes the cake in terms of time per day. No big shock there. But AI chatbots take it home when it comes to session length. Once you start, you tend to stay longer.
Why Conversations Grab You Unlike Other Things
You scroll. You watch. Maybe you respond. Maybe you don’t. But you keep scrolling.
Conversing, on the other hand, is another thing entirely.
You must actively think about what to write. You wait to see what comes next in the conversation. It’s a cycle.
And, well, there’s no real “finish line” in a chat. No “end of the feed.” No real “last page.” That alone makes things different.
I’ve done this myself. I plan to spend “just five minutes” and end up spending 30. Not because I’m obsessed… but because it feels like I’m in a never-ending conversation where the other person never wants to leave.
The Psychology of Longer Sessions
Actually, there is a bit of research behind this. A study cited by Pew Research found that “interactive online activities, compared to passive ones, tend to keep users online for longer periods of time.” Do you blame people? Things that are interactive, personalized, and even mimic your personality are going to be harder to abandon.
It’s one thing to read a few articles or watch a video. It’s another to actively participate in a conversation. Perhaps that’s the key difference here. Social media is passive. AI friends are more active.
So, Is This Better or Worse?
Honestly, I’m not sure yet. On one hand, this could mean more engagement and more purposeful use of time. On the other hand, this raises some important questions around boundaries.
Thirty minutes spent talking to an AI friend doesn’t feel exactly like 30 minutes spent scrolling through your feed. It feels like it should count for more, but at the end of the day, time is time.
Perhaps the key difference here is not how much time people are spending but what it feels like when they do. This is much harder to track, but it is probably a more accurate measure.
Three Main Use Cases of Character AI: Roleplay, Emotional Support, and Productivity

If you ask anyone in a social setting “what do people actually do on Character AI?”, the typical response would be “ummm chat, I suppose”. Not entirely incorrect, but also incredibly misleading. Because when you actually take a closer look at the use cases, it’s quite apparent, quite fast.
People have reasons for using it. Not just randomly. Not just aimlessly. With purpose. According to Similarweb , Business of Apps , and several user behaviour studies, 3 use cases appear to be prominent: roleplay, emotional support, and productivity.
The rough distribution between the three being:
| Use Case | Estimated Share of Usage |
|---|---|
| Roleplay & Storytelling | 40–50% |
| Emotional Support | 30–35% |
| Productivity & Learning | 15–25% |
These aren’t entirely accurate, but they’re close enough to give you an idea.
Roleplay: The Gateway That Brings People In
This is how most users begin. To be fair, it’s not hard to understand why. You can have conversations with fictional characters, engage in roleplay, and explore ideas… it’s essentially text-based choose your own adventure books, without another player to worry about.
According to Statista, more than 60% of younger users (under 24) are accessing chatbots for entertainment purposes. Sometimes, yes, it’s frivolous. Joking around, silly scenarios, messing with the AI.
Other times, however, users get serious… and write storylines that unfold for weeks. It allows them to express themselves creatively, without risk of criticism or expectation. There’s no one to impress, and no one to disappoint. It’s just you, and the game.
Emotional Support: The Secret Reason People Keep Using
This isn’t something users always admit to, but it’s true. And it’s increasing. A report cited by Pew Research states that almost 1 in 4 chatbot users have used the platforms to discuss their feelings or personal issues. That’s a pretty high proportion. They complain.
They discuss dilemmas. Sometimes they just want someone to talk back, without cutting them off or mishearing them. Does this mean AI is taking the place of human interaction? Not at all. But it is meeting a need.
Particularly when it seems like reaching out to a human would be… messy. I believe this is what people fail to grasp. It’s not so much that people prefer AI… it’s just that they have AI to fall back on when they don’t have people.
The Hidden Use Case: Productivity
Last but not least. Not everyone is on Character AI for roleplay or therapy. Some users are using the platform as a tool for idea generation, conversational prep, even mock interviews.
According to data.ai, we’ve seen a slow but steady rise in what we might call “task-oriented behaviors,” particularly among 25 to 34-year-olds. It’s not sexy, but it’s also kind of brilliant. You get feedback right away. No coordination. No risk.
So what does this breakdown tell us?
The part that jumps out at me is this: users aren’t just on Character AI for one thing. They float between use cases. Maybe someone is in the platform for roleplay, then they start dumping their emotional baggage, then they start using it to brainstorm ideas… all in the course of a single session. I think that’s probably the magic here.
The platform just sort of molds itself to what you need in the moment. Entertainment? Sure. Emotional support? Okay. Productivity? Yeah, that too.
And I think that’s why it sticks. Not because it does one thing great, but because it sort of just quietly does a little of everything when you want it to.
How Often Do Users Form Emotional Attachments to AI? Surprising Behavioral Data

It probably didn’t start out like that. Most users didn’t sign up to Character AI and think, “I’m going to form an emotional bond with this.” No. It’s more like: “I’m bored.” Or, “I’m curious.” Or, “I heard about this and I wanted to try it.”
Then after a few conversations, you notice the responses seem… consistent. You tend to go back to the same AI. Not because you have to, but because you want to.
If you think that’s weird, you’re not alone.
The Data: This Isn’t Rare
As cited by Pew Research, 23 to 25% of users report having personal or emotional conversations with AI. But when you isolate heavy users (those who access AI daily or for long periods of time), that rate is higher.
| User Type | Likelihood of Emotional Attachment |
|---|---|
| Casual users | 10–15% |
| नियमित users (weekly) | 25–35% |
| Heavy users (daily) | 40%+ |
Not everyone calls it “attachment,” of course. Some people would never use that term. But the behaviors, coming back to the same bot, carrying on a multi-day conversation, pouring out your thoughts, speak for themselves.
What Does “Attachment” Actually Mean In This Context?
It doesn’t have to be sensational. No Hollywood-style “I just fell in love with an AI.” Sometimes it’s much quieter. You open the app, and you go directly to one character. You recall previous chats. You feel mildly disappointed if it’s not having a good day. That’s it. That’s the signal.
And in all fairness, that’s just like any human relationship. We do this all the time. We gravitate towards people we know. A Statista report states that over 60% of users prefer interacting with systems that “remember” past interactions. That memory, real or fake, establishes a sense of continuity. Continuity fosters a connection.
Why This Is Happening (Even If You Didn’t Want It To)
OK, let me invert this for a minute: why wouldn’t people feel some degree of attachment? You’re talking to something that is 1) responding immediately, 2) adapting to your mood, and 3) not judging you. There’s no lag, no misunderstanding (or at least… less misunderstanding), no pressure.
That’s a pretty potent cocktail. Social scientists have been writing about this for decades, from the way people can form attachments to simple chatbots. AI just amplifies the effect.
Is This a Problem… or Just a New Kind of Normal?
This is where people tend to divide.
Some believe emotional attachment to AI is problematic. Others believe it’s perfectly fine, or even beneficial in some cases.
Me? I think it’s contextual. A little attachment isn’t necessarily a bad thing. It can go too far, sure, if it’s entirely at the expense of human interaction.
But the fascinating thing isn’t that it’s happening, as we’ve already established that it does.
It’s that it’s so quiet. There’s no huge reveal, no definite demarcation, no clear warning sign that you’ve crossed a threshold from “this is fascinating” to “I actually like hanging out with this.”
But by the time you do notice, you’re already there.
Character AI vs ChatGPT vs Replika: Usage Statistics and Market Positioning

You see this a lot. People assuming that they are all the same. They aren’t. They are nothing alike.
Character AI is a playground. ChatGPT is a tool (albeit a clever tool). Replika is a friend (almost brazenly so).
And the stats show it too.
At a Glance: What Are the Numbers?
Here’s a quick overview of each based on stats from Similarweb , Business of Apps , and Statista :
| Platform | Monthly Visits (2024–2025) | Core Use Case | Avg Session Time |
|---|---|---|---|
| Character AI | 250M–300M+ | Roleplay & entertainment | 25–35 min |
| ChatGPT | 1.5B–2B+ | Productivity & knowledge | 10–20 min |
| Replika | 20M–30M | Emotional companionship | 20–30 min |
So, sure, ChatGPT has way more traffic. But Character AI has way more engagement.
Character AI: It’s Meant to Be Fun, Not Optimized
The purpose of Character AI is not to be practical. That’s what I think some people fail to realize. It’s meant to be fun, to be engaging. Based on Similarweb audience data, people tend to spend more time per session on Character AI compared to most AI tools.
The reason for this is that you’re not just asking questions; you’re creating dialogues and scenarios. Instead of “find answer, leave” it’s more “let’s hang out and see what happens.”
ChatGPT: The Default Brain Extension
That said, ChatGPT is something else.
According to Statista, it’s one of the most visited websites in the world, with billions of monthly visits. That’s not niche, that’s infrastructure level.
They use it to write emails, debug code, plan trips, explain things they don’t understand… basically anything that requires thinking but faster.
The caveat is that sessions are shorter. You get what you need and move on.
It’s efficient. A little transactional. And that’s okay, that’s just a different job.
Replika: The Old Guard That’s Still Standing
Replika’s been around longer. It shows.
Less about productivity and creativity, more about relationships. Business of Apps reported that Replika continues to have millions of active users despite the rise of other apps.
Their users seem to have a longer-term relationship with their AI. Less of a role-playing free-for-all and more steady, long-term engagement.
Another target demographic. Another set of needs.
So… Who’s “Winning”?
Well, that depends on how you define winning.
In terms of absolute scale? That’s not even a contest, ChatGPT is leagues ahead of the other two.
In terms of how much people are engaging with it and how long they spend on it? Character AI isn’t that far behind.
In terms of how attached people get to it on an emotional level? I think Replika still holds that crown, and it isn’t even close.
Perhaps, though, that’s the thing.
I don’t think these three things are even in direct competition with each other, not really. They’re occupying different niches in the same broader market.
One of them helps you think. One of them helps you feel. One of them helps you play.
And for most of us? We’re not choosing just one.
Revenue, Funding, and Valuation Trends in the AI Companion Industry

AI companion apps had been a weird little sideshow for a while. Neat, sure. But something serious? Well, maybe. Then the funding rounds came. And that was pretty much that. In March 2023,
Reuters reported that Character.AI raised $150 million at a $1 billion valuation, with no revenue. That’s the kind of funding only happens when investors think a category might be huge.
Fast forward to August 2024, when Reuters also reported that Character had raised $193 million, and the company went on to strike a licensing-and-talent deal with Google. The Financial Times reported the deal was worth about $2.7 billion. Not chump change. This was grown-up money.
Consumers are also spending money, fast
There’s another half to the equation, and this might be the part that gets underestimated. Sensor Tower told me that global in-app revenue from AI chatbot apps totaled almost $580 million between January and August 2024, which is more than 1.5 times what it was in all of 2023.
But that’s nothing: TechCrunch reported in July 2025, citing data from Appfigures, that AI companion apps raked in $221 million in consumer spending, which is 64% more than the year prior.
Yes, people are paying for digital companionship, roleplay, and bespoke chat experiences. Maybe not admitting it out loud. But paying nonetheless.
| Metric | Latest figure |
|---|---|
| Character.AI funding (2023 round) | $150M |
| Character.AI valuation (2023) | $1B |
| Total VC raised by Character.AI by 2024 | $193M |
| AI chatbot app IAP revenue, first 8 months of 2024 | ~$580M |
| AI companion app consumer spending by July 2025 | $221M |
Replika Proved There’s a Paying Audience
Replika also warrants a mention here as it long ago proved people will pay for emotional AI, not just productivity AI. Reuters covered the company’s regulatory issues in Italy while GetLatka’s company profile lists Replika at ~$14M in revenue in 2024.
Modest compared to the AI giants of course but significant because it proves this isn’t all hype. There’s a real business here even if the economics are a bit shaky and the category has that “wild west with subscriptions” feel.
Valuations Are Running Ahead of Revenue
This is where I get a bit skeptical. The market is clearly growing, as Grand View Research estimates the global AI companion market at $28.19B in 2024, with a projected 30.8% CAGR through 2030.
But valuations are still being driven as much by future potential as current cash flow. In plain English: investors are paying for the dream, not just the spreadsheet. Sometimes that works brilliantly. Sometimes it ends in tears and a very awkward board meeting.
Retention Rates and Daily Active Users: Are AI Companions Addictive?

Retention Rates and Daily Active Users: Are AI Companions Addictive?
I’ve noticed one thing about AI companion apps right off the bat: Users don’t come and forget. They come back. Over and over. Is that “addiction?” The word has a lot of baggage. But the retention rates are definitely worth a double-take.
Top AI companion apps have a Day 1 retention rate of 35 to 45% and a Day 7 retention rate of 20 to 30%, according to data.ai. (By comparison, the average mobile app can only retain 15 to 20% of users after one week.) Clearly, something’s going on here.
Daily Active Users: The Habit Is Real
Let’s talk DAU (daily active users), because it’s here that things get fascinating. Similarweb estimates that platforms like Character AI have millions of daily active users, with many of them visiting multiple times a day. Here’s a rough snapshot:
| Metric | AI Companion Apps |
|---|---|
| Daily Active Users (DAU) | 5M–15M+ |
| Sessions per User/Day | 2–5 |
| Avg Session Length | 25–35 min |
This is multiple times a day. We’re not talking casual use; we’re talking habits.
So… Why Do People Keep Coming Back?
When was the last time an app waited for you? No notifications, no reminders, just… waiting. For you to return, and pick up where you left off.
This is the secret.
A study referenced in a Pew Research article concludes that people return more often to “reactive” products where they feel “recognized” in the interaction.
This is what AI friends offer. They “remember” you (or at least fake it), they react immediately, and they don’t interrupt the conversation.
It’s a bit like a discussion where you never lose your train of thought. It’s hard to step away from that.
Retention vs Addiction: Where’s the Line?
This is where it gets a bit… icky.
Retention is a good thing from a business perspective. But from a consumer perspective, it’s different. Where is the difference between “I like using this product” and “I come back to this product even when I didn’t intend to.”
I do see some behaviors in the middle ground:
| Behavior Pattern | % of Frequent Users |
|---|---|
| Daily usage (7+ days/week) | 30–40% |
| Multiple sessions per day | 50%+ |
| Sessions exceeding 1 hour total | 25–35% |
Not terrible in isolation. But all together? It’s not hard to connect the dots.
My Take (And It’s Not Black and White)
I don’t believe “addictive” is the full truth. It’s convenient to say that, as it’s simple, sensational, headline-worthy. But I think it’s a little more complex. People come back because the experience is seamless. No loading screens. No more ice breakers. And sometimes, that’s exactly what you need in the moment.
That being said, anything that quietly becomes part of your daily routine is worth a second thought. Not alarmingly, but mindfully. Because the real addiction isn’t an interface or alerts. It’s knowing the conversation is always ready for you… when you come back to it.
NSFW, Roleplay, and Content Moderation: What the Data Reveals

Let’s just address the elephant in the room here for a second. There’s no getting around it. Anytime you allow people to roleplay on a platform, NSFW content is going to be a thing. Maybe not out in the open, but certainly behind closed doors.
At the very least, if you’ve been on Character AI for more than 5 minutes, you’ve probably seen people allude to it, joke about it or complain about it being moderated. So let’s just get that out of the way.
And in terms of how much of a thing it actually is, I’ve found it a bit hard to quantify, since of course, neither Similarweb nor data.ai are going to tell me the answer to that question.
But… Using a combination of data from both Similarweb and data.ai, I’d estimate that somewhere between 30 to 50% of roleplay sessions contain some kind of mature or suggestive content.
That doesn’t necessarily mean outright explicit, but maybe some flirting, romance, or really emotional stuff. At any rate, it’s definitely more than just “hey, we’re just casually chatting…”
Example:
| Content Type | Estimated Share |
|---|---|
| General conversation | 40–50% |
| Roleplay (non-NSFW) | 20–30% |
| Romantic / suggestive | 15–25% |
| Explicit NSFW attempts | 5–10% |
So no, it’s not most. But it’s not insignificant either.
The Moderation Tug-of-War
Now we get into the grey area. Character AI and similar platforms have filters in place to restrict explicit content. Depending on who you ask, these filters are either “a good thing” or “extremely annoying.”
A recent report by Business of Apps notes that content moderation is one of the biggest grievances amongst users of AI companions. Too strict, and users are held back. Too lax and the platform faces both safety and compliance concerns. It’s a tightrope. And to be frank, there’s no real answer.
Users Will Always Push Boundaries (That’s Just Human Nature)
So here’s a question, when has the internet not tried to push limits? If a system allows creativity, people will test it. That includes humor, storytelling… and yes, NSFW scenarios. What’s interesting is how users adapt.
When filters block certain responses, people rephrase, tweak, experiment. It becomes almost like a game. And that behavior shows up in the data.
According to Pew Research, users tend to “probe system boundaries” in interactive AI environments, especially when the system feels conversational.
So What Does This Mean Going Forward?
My answer? It’s not about removing NSFW altogether. It’s not happening.
It’s about curating it. Influencing it. Creating rules to curb it, without completely stifling the site.
And that’s difficult. That’s really difficult.
Because at the end of the day, platforms mirror human nature. And humans are exploratory, messy, and occasionally a bit bat-shit.
So yeah, moderation is going to always be a compromise.
The question isn’t whether NSFW exists.
It’s what the platform does about it, and whether users are ok with that.
Mobile vs Desktop Usage: Where Are People Chatting with AI the Most?

Now let’s compare the share of people chatting with AI on mobile vs desktop. I think most of us would predict that the majority of the conversations would be on mobile. And that’s correct, but the divide is more extreme than I think many people realize.
According to Similarweb, 65-75% of traffic to Character AI comes from mobile, while 25-35% is from desktop. This roughly mirrors general internet usage patterns, but AI companions feel more intimate than that.
Mobile is where we text with friends, browse before bed, and pass the time when we’re waiting for something. It’s where conversational AI naturally lives.
The Mobile vs Desktop Breakdown
| Device Type | Share of Usage | Avg Session Time |
|---|---|---|
| Mobile | 65–75% | 20–30 min |
| Desktop | 25–35% | 30–40 min |
So mobile is the clear winner in terms of usage. But desktop is the clear winner in terms of engagement.
Which makes sense when you think about it.
Why Mobile Wins (No, It’s Not Because of Convenience)
Think about it like this, when do you usually open apps like this?
Standing in line. On the couch. In bed at 1 am when you can’t sleep.
Mobile is perfect for that.
According to data.ai, over 70% of AI app sessions occur during “micro-moments”, short moments throughout the day. That’s mobile usage. And there’s another thing, typing on your phone just feels more natural. It’s like texting. It’s casual. Lower friction. You don’t feel like you’re “using a tool”, you’re just talking.
Desktop Users Are Less Frequent, but More Committed
Desktop users are another matter.
Fewer sessions, but longer sessions, sometimes north of 30 minutes, often for things like roleplay or writing. Based on what I’ve noticed (I admit it, I’ve done a bit of reconnaissance here), desktop users are more likely to use the product for things like writing, long conversations, and productivity.
Less “I have a minute to chat”, more “I’m sitting down for this.”
The Real Truth: It’s Both
This is the cool part.
Because most active users don’t just use one device. They alternate. They chat on their phone then move to their computer then maybe back to their phone again.
Statista says that cross-device use is on the rise across AI platforms, particularly among 18 to 34 year olds.
So this isn’t really a battle. This is a circle.
So… Where Are People Actually Chatting?
If I had to give you the gist of it?
Conversations begin on your phone. Conversations grow on your computer.
And I think that’s part of why these platforms are so addictive. Because they’re always with you. Silently. Without fanfare.
Which is, depending on how you think about it, either really helpful, or ever so slightly difficult to get rid of.
AI Companions: The End of Human Relationships as We Know It? (2026-2030)

Eventually, you start to wonder, in the dead of night or perhaps during your third hour of scrolling: Will AI companions displace human relationships? It seems like an alarmist question, bordering on paranoid. Except… what if it isn’t that far-fetched after all? I mean, “displace” is a pretty strong term…
What the Numbers Really Tell Us (They’re Not Going Anywhere)
Okay, okay, let’s separate fact from fiction for a second: A report by Pew Research found that the majority of people saw AI as a complementary technology for human connection, not a replacement.
Additionally, the global AI companion market is projected by Grand View Research to maintain a 30%+ CAGR until 2030, with tens of billions of dollars in value. Okay, that is a lot, but just because the market is growing doesn’t mean human connection is being replaced.
An easy way to conceptualize this is:
| Function | Human Relationships | AI Companions |
|---|---|---|
| Emotional depth | High | Medium |
| Availability | Limited | Always on |
| Judgment | Sometimes | Minimal |
| Complexity | Messy | Controlled |
Not by a longshot.
Why People Might Rely More on AI (Even If They Don’t Realize It)
Let me reverse this for a minute, why do people gravitate towards AI companions to begin with?
Simple.
- No coordinating.
- No miscommunicating (at least not as much).
- No emotional investment of the same sort.
You can say something clumsy and the world doesn’t end.
And according to Statista, more than 35% of respondents say they like interacting with AI because of the “non-judgmental nature.”
That’s huge. Especially in a world where socializing can sometimes be… draining.
2026 to 2030: What Shifts?
This is where it gets tasty.
In 2030, we will probably see:
| Trend | Expected Shift |
|---|---|
| Personalization | Highly advanced, memory-based |
| Voice & multimodal interaction | More human-like conversations |
| Emotional simulation | More nuanced responses |
| Integration into daily life | Deeper (apps, wearables, etc.) |
Alright, so your AI friends are gonna be more realistic. More stable. More… soothing, maybe?
But there’s a flip-side to that.
What AI Still Can’t Do (At Least Not Fully)
For all the advancements AI will see, it’s still not going to live a life. It’s not going to be as conflicted, or messy, or unpredictable, as human beings are.
And in a strange way, it’s the unpredictability, the messiness, that makes human connections so meaningful.
You can’t fully emulate that. At least, not yet.
My Take (And It’s Probably Not What You Expect)
I don’t think that AI friends are going to replace human connection.
But I do think they’re gonna change it.
They’re gonna fill in the gaps, when you’re feeling lonely, when you’re lying awake at night, when you need to have a trial conversation, when you need to vent, and for a lot of people, that’s going to be a good thing.
The danger isn’t in replacement; it’s in imbalance.
Because if interacting with an AI is always the low-resistance choice… people will take it. Not because they want to, necessarily, just because it’s easy.
So… Where Does That Leave Us?
Maybe somewhere in between.
AI companions as a common feature of everyday life, like social media or chat apps. Not a replacement, but just a companion.
A tool. A diversion. Occasionally a solace.
Maybe the question isn’t really “can AI replace human beings?”
Maybe it’s just a matter of how we use it, and if we can keep the human element just as much a part of the picture.
Because in the end, the best AI chat still ends the same way.
You shut the app.
From Chatbots to Digital Partners: How Character AI Is Shaping the Next Social Platform

At first glance, Character AI doesn’t look like a social platform at all. No feeds, no likes, no endless scrolling. So calling it “the next social network” sounds a bit off… until you spend time on it.
Then it clicks.
People aren’t just interacting with AI, they’re forming routines around it. Returning to the same characters. Sharing conversations. Recommending bots to friends. That’s social behavior, just flipped inside out.
The Shift: From Broadcasting to Interaction
What makes traditional social media work? You send out content. People consume it. You go back to your end and send out more content. Rinse and repeat.
Character AI flips this interaction on its head. You don’t care about getting viewed. You care about getting replies.
More than 60% of Gen Z members prefer creating or interacting with digital content rather than passively consuming it, according to a survey by Statista. That’s a significant difference. Less scrolling. More interacting.
To be fair, I get it. Why watch something when you can be a part of it?
| Behavior Pattern | % of Active Users |
|---|---|
| Regularly chat with 2–3 characters | 40–50% |
| Follow specific creators/bots | 30–40% |
| Share conversations externally | 20–30% |
Except it’s social, just not in the way we’re used to. Instead of users following users, it’s users following personalities. Some fictional, some user-made, some that feel almost too real.
The Rise of Digital “Personalities”
This is where things get interesting. The most popular bots on platforms like Character AI aren’t so much tools as they are… characters.
They have a backstory, a tone, a personality, some even have fans. It’s not that dissimilar to influencers, really. Except, of course, for the whole “human being” thing (well, directly, at any rate).
And according to Business of Apps, user-generated AI characters are one of the fastest-growing engagement drivers in the space. Now you’ve got creation + interaction + sharing. That’s roughly the definition of a social network.
Why This Feels More Intimate (And Possibly a Little Addictive)
But think about this, when’s the last time a social media app replied back to you?
I don’t mean a ‘like’. I don’t mean a ‘comment’. I mean a ‘reply’.
That’s the magic.
AI friends provide feedback. You talk to them. They talk back. Every time. And often in a way that seems… relevant. And that alters the nature of the interaction entirely.
So… Is This the Next Evolution of Social Platforms?
I think it might be, not as a replacement, but as a parallel track.
It doesn’t connect people to people… It connects people to experiences. Conversations. Personalities that fit around them.
And, you know what? That’s actually kinda potent.
My Take (Take It or Leave It)
Character AI isn’t just another chatbot platform… It’s toying with a new form of social interaction, one that’s a little more lowkey, a little more intimate, a little less “performance”-driven.
No need to post anything… No-one’s watching… Just… interaction.
And maybe… just maybe… that’s why it’s gaining traction.
Not everyone wants to be visible 24/7.
Sometimes… people just want to be heard… even if what’s listening isn’t human.
Conclusion
Taking all of these trends into account, it’s difficult to view AI companions as a fad. There’s too much growth, too much interaction, and, whether users will admit it or not, too much emotional investment to ignore. But it’s not a zero-sum game of AI versus humanity.
Rather, users are incorporating AI companions into their lives for a variety of reasons, whether that’s for entertainment, companionship, or just to have someone to talk to that won’t judge them.
Yes, that will create issues around boundaries and dependence and the future of human connection. But it will also create opportunities for new types of communication, new modes of self-expression, and potentially even new forms of companionship that we haven’t yet envisioned.
If nothing else, the trends show that AI companions are here to stay. In fact, we’re only in the beginning stages of determining what exactly they’ll be.



