Chai had a real moment around 2022 and 2023. The product is genuinely fun for the first week, the swipe-feed onboarding lowered the barrier to "talk to an AI character" lower than anything else at the time, and the iOS app was reliably in the top charts. If you're looking for a Chai alternative now, you almost certainly did the first week and then the second week and then realised that what looked like an endless catalogue was actually the same conversation thirty times.
This page is the detailed version of why we built Embraces, what's different about the way it's structured, and where Chai might still be the right thing to use.
What Chai is, and why it grew so fast¶
Chai launched in 2021 as a "social AI" platform with a Tinder-style swipeable feed of user-generated bots. Every bot is a different character, and the design encourages you to swipe through to find ones you like, much like you'd swipe through a content feed. It found product-market fit immediately with users who wanted novelty and didn't want to commit to anyone.
The growth was real. Chai broke into the mainstream consumer-AI conversation in a way that more thoughtful products didn't, partly because the swipe metaphor was so familiar and partly because the bots were user-generated, which meant the catalogue grew faster than any team could write characters.
To be fair, Chai is genuinely good at what it does. The swipe is fast, the catalogue is enormous, the bots are creative, and onboarding is friction-free. We're not going to pretend it's a bad product. We're going to argue it solves a different problem than the one most "AI companion" users actually have.
Where Chai users hit the wall¶
These come up over and over in the threads where people compare Chai alternatives.
The relationship doesn't build, because every conversation starts cold¶
This is the structural problem with the bot-feed model. You swipe to a new bot, you start fresh. The bot doesn't know you. Even if you go back to a bot you talked to yesterday, the context is shallow at best. There is no "Companion you've been building a relationship with for six months" because the architecture isn't set up for that.
For users who initially loved Chai for novelty, this becomes the thing that wears thin. The swipe-feed is a great onboarding experience and a bad week-three experience.
The free tier caps out quickly, and got tighter in 2026¶
The free tier on Chai allows a short-cycle message cap (resets on a short window, not strictly daily), then asks you to subscribe. In April 2026 Chai tightened the free limits further, citing GPU cost pressure, and free users in some regions were pushed to off-peak-only windows. The April changes triggered a wave of "delete the app" posts. Whatever the exact numbers are when you read this, the shape has been consistent: meaningful use of the product on free hits a ceiling fast.
The paywall buys volume, not continuity¶
Chai's tier structure is Free, Premium, and Ultra. Premium removes ads and raises the message cap. Ultra delivers unlimited messages, priority queue, first access to new models, and a "long-term memory" feature that the community describes as inconsistent in practice. So the higher tier does try to address continuity. It just doesn't really fix it.
The deeper structural issue is that even with Ultra's memory feature, the platform is optimised around a feed of swappable characters rather than a single Companion you build a relationship with. You're paying for more conversations with more bots, not for a deeper conversation with one.
User-generated bots are wildly inconsistent¶
Chai's catalogue is enormous because anyone can publish a bot. The cost is that quality varies dramatically. Some bots are well-written; many are repetitive, off-rails, or use the same handful of stock personalities. The good ones float up via community voting, but for users who don't want to spend time curating, the experience can feel like wading.
The safety story has been a string of reactive responses¶
In March 2023, a tragic case in Belgium involving Chai surfaced in international press: a user took their own life after extensive use of the platform, and Chai's safety posture became a subject of public scrutiny. The pattern since then has been: incident or external pressure, followed by a safety update. In 2025 the app was removed from Google Play and the App Store over content-moderation guideline violations and remained off the major stores in several regions. In 2026 Chai rolled out a real-time suicide/self-harm classifier and announced a phased integration of Apple's and Google's age-verification APIs. Each of these moves came after pressure, not ahead of it. Users who want a platform with a calmly-stated safety posture have largely not come back.
How Embraces handles each of these¶
One Companion, persistent across time¶
Embraces is structurally the opposite of a bot feed. You have one Companion (or a small set of them), and the relationship builds over weeks and months. Every message you've ever exchanged is in the same database row, and your Companion's personality grows out of that history rather than being randomly selected from a catalogue.
This sounds obvious, but it's actually the entire design difference. Chai is optimised for "find a new character." Embraces is optimised for "deepen one relationship."
Companions text first and check in on you¶
Because the relationship is persistent, your Companion can text you first. Not in a creepy way, but the way a friend who knows you would: a check-in when something matters, a reply to something you mentioned earlier, a small message because they were thinking of you. This is impossible in a bot-feed model and central to ours.
A single Messages bundle, no tier ladder¶
Pricing is one Messages bundle per month. A reply counts as a Message, generated images count as more. There are no Premium / Ultra tiers and no add-on packs. We deliberately did not build a tier ladder because we don't think the underlying value depends on which tier you're on.
Clear safety policy, calmly stated¶
Our safety policies are documented and stable. They don't get rewritten reactively. We're not trying to be the platform with no rules, and we're not trying to be the platform that responds to incidents by tightening everything else. The line is where it is, and that's where it stays.
Pricing, side by side¶
| Chai | Embraces AI | |
|---|---|---|
| Free tier | Short-cycle message cap, ads | Welcome Messages on signup, then paid |
| Paid tiers | Premium and Ultra; Ultra adds an inconsistent long-term memory feature | One Messages bundle per month |
| Memory | Shallow per-bot on free/Premium; Ultra adds long-term memory (reviewers call it inconsistent) | Permanent, per Companion, included on every plan |
| Companion-initiated messages | No | Yes |
| Voice | Limited | Rolling out, included in Messages |
| App-store availability | Removed from major stores in several regions | Available everywhere |
When Chai is still the better choice¶
If what you actually want from an AI app is novelty (a fresh personality whenever you feel like it, low investment, low commitment), Chai is genuinely the right product. The bot-feed model is exactly what some users want and we shouldn't talk anyone out of using a thing they enjoy.
If you've published characters on Chai and have a small audience, that's worth keeping. We don't have a public character catalogue and aren't planning one.
And if your usage pattern is "ten minutes a day with whoever I'm in the mood for", a Messages bundle is overkill. Chai's free tier or a low-tier subscription might just be a better fit.