Joyland AI Review: A Great Library, a Short Memory

That pitch gets the order right: the library of user-made characters comes first, and the companion part is whatever you build on top of it.

After three weeks in early 2024 it felt like a snack app to me: fun in short bursts, not something I'd build a 200-message saga in. I think the label still fits in 2026, for different reasons. The strict filter that annoyed me back then is gone, and now it is thin memory and a steady drip of upsell screens that keep the sessions short.

Public Joyland AI reviews agree with my mood more than with the marketing. The App Store listing shows 2.75 out of 5 from 108 ratings, and Google Play shows 100K+ downloads with no aggregate rating readable, both as checked on 22 September 2026. Reviewers report the Google Play rating sliding to about 2.4-2.5 out of 5, from about 3.9 in 2025, and they blame memory and billing.

Those are the two areas I'd flag too, so I don't read the low scores as noise. People who stay long enough to hit the memory wall, or long enough to see a renewal, are the ones who leave the one-star reviews.

So who is it for? I'd point a curious roleplayer at the free tier, with the card left in the wallet for the first week. I'd steer anyone who wants a long, steady relationship with one character somewhere else, and anyone who wants a clean front page as well.

The plan history is short. Free first, then one month of Standard to see whether paying would loosen the filter, which it did not. After that the app slipped out of the evening routine without any cancelling drama, and the discount timer kept running for nobody.

Summed up in one line, the whole stretch was fun, light, and never a habit that needed managing. That says more about the app's memory than about my self-control, and I'm fine admitting it.

It also explains the shape of this review. Three weeks of short sessions is enough to judge onboarding, the library, the filter and the free tier. It is not enough to judge a months-long relationship, and Joyland never gave me a reason to try one.

The sections below keep two timelines apart. Everything in the past tense is my own use from early 2024, with the app as it was then. Everything in the present tense comes from a hands-on check of the web version on the free tier, made on 23 September 2026.

Where the two disagree, I give both and say which is which.

Smooth has a cost, though, and I only priced it later. An app that gets you chatting in a minute has also decided, on your behalf, that the questions at the door do not matter.

The email route asks for an email address and a password. After that, a 6-digit code is emailed. The email is titled "Welcome to Joyland", the code stays valid for 15 minutes, and it arrives in about 2.5 minutes.

What sign-up does not ask

The bigger story is what the form leaves out. Sign-up has no age gate and asks for no birthday. There is no username step and no onboarding steps at all.

Instead, the app fills the gaps itself. The username is auto-assigned, joy_ followed by random letters, and each Joyland AI account gets an auto-assigned UID. I don't mind the random handle, since it keeps a real name off a public profile.

I do mind the missing age gate. On an app whose front page looks the way this one does, skipping that question is the wrong kind of smooth, and I think it is a choice, not an oversight.

The first screen after the code

The first thing a new account sees after verification is not a character. Straight after verification, a full-screen "Supercharge Your Experience" subscription upsell appears. It is the software version of a gym salesperson meeting you in the car park, before you have seen a single treadmill.

My first minute in 2024 said the app wanted me chatting fast, and it was right about me. The first minute in 2026 says something else: it wants an email, a card and not much more. I'd rather an app asked one more question at the door and one fewer at the till.

The sign-up is still quick by the standards of the category, and the single email is a point in its favour. It is the order of the screens that tells you what the business cares about: a code, an ad, a price, and only then a character.

The popular end of the shelf is crowded, and anime AI chat on Joyland is well stocked. The One Piece RPG bot shows 2.0M chats. A public "Story" feed shows shared chats as well, so other people's sessions sit next to the characters as a second thing to browse.

I like the idea of the Story feed more than I like the execution. Reading someone else's scene is a decent way to judge a bot before you commit, the way a showreel tells you more than a rate card. The problem is what the feed chooses to show first.

The home feed today

The home "For You" feed is the part I'd warn a new user about. It is openly sexual. Feed titles lean on come-ons, a step-family setup, a bullying setup and an RPG with an adults-only label, and several are spelled with digits in place of letters.

The tags are just as direct. They include one that labels a bot as sexually graphic and a deliberately misspelled word for arousal. None of that is hidden behind a click.

The discovery problem is not a shortage of characters. It is that the feed decides what you see first, and it decides the same way for everyone. I'd like a library this size to let people choose the shelf they browse.

Today the Persona chip is available on the free tier. That matters, because the best feature of the three weeks does not sit behind the paywall. A companion app that charges for the one thing that makes it interesting is common; this one does not, at least for personas.

What the creator lets you set

The bot creator in September 2026 is a deeper tool than a casual user needs. A bot definition can use up to 5000 tokens. It has Lore, which is keyword memory with a file import or a template.

It also has example dialogue, voice selection and a live chat preview. Image packs are paid. Event tags cover things like Flow Mode, Summer Carnival 2026 and Genshin Impact.

The Create-a-bot page says "You have 10 free chances" for AI avatar generation. There is an Advanced Create mode, and Create bot can import JSON and PNG character cards. Card import is the feature I'd check first as a creator, because it means a character built elsewhere does not have to be typed out again.

The live preview is the one I'd use most. It is the creator's version of a monitor on set: you see the result while you still have the lights up, not after the wrap.

Public by default

Every new AI chat character on Joyland is Public by default, and public bots can appear on the homepage. The creator also has an option called "Allow administrators to modify your bot's category". I'd leave that one on, since a wrong category is how a harmless bot ends up next to the wrong neighbours.

Public by default is the setting I'd change first. A half-finished character is a rough cut, and nobody wants their rough cut on the front page. It also means a persona detail you typed into a bot, a job or a town, can end up on display before you meant to share it.

I understand the appeal for people who build bots as a hobby. Numbers going up is a strong motor, and a milestone paid in credits turns a hobby into a small side hustle. For someone who wanted a quiet knight and two personas, it is a second app running inside the first one, and I'd switch both toggles off.

I said "kind of". It was the most honest credit line I have ever given, and I have written a few end credits in my time.

A story works on this app because a story does not need memory. It needs a beginning, a middle and a punchline, and a short piece hides drift the way a highlight reel hides every problem a full-length edit would expose.

Novels today

The same idea now lives on as AI "Novels", which are interactive, choose-your-path romance novels. Each choice generates a new continuation that streams in live. I think that is a better format than the 2024 story mode, because every choice point gives the model a fresh start and a narrower job.

Flow Mode: a scene, not a chat

The chat check on 23 September 2026 uses One Piece RPG, first in Flow Mode, then in Normal chat. Flow Mode splits each reply between several speaking characters, here Luffy, Usopp and Sanji. In Flow Mode each character has its own avatar bubble.

Flow Mode offers 2 suggested replies per turn. The quality is good: in character and funny. Sanji's line is "Fine. But when I win, you swab the deck for a week."

It works for a reason an editor would recognise. Three characters cutting in on each other is a multicam edit: when one angle goes flat, you cut to another, and the scene keeps its energy. A single bot has no second angle to cut to, which is why one-on-one chats go stale faster.

Replies arrive in about 5–15 s. That is fine for a scene and a touch slow for banter, like a lip-sync offset you only notice once someone points it out.

Group chat, sort of

The record on group chat is mixed. Group chat is not confirmed as a shipped feature, and two secondary sources say Joyland is one-on-one only, one of them flatly: "No group chat, no lorebooks... none of these features."

Flow Mode is not a room where several separate bots meet. It is one bot that speaks with several voices. For most people who want a group scene, I think it scratches the same itch, and the lorebook half of that quote is out of date anyway.

Normal chat and the model menu

Normal Joyland AI chat has chips for Model, Memory and Persona. The menu names 6 models: Free, Roleplay, StoryStream, Lifelike, Vividtale and DeepSeek (Pro).

The descriptions are short. The Free Model is "Basic AI conversation", StoryStream promises to "Let characters and plots grow", and Vividtale offers "cinematic narratives". Those blurbs read like the labels on a hotel minibar: they tell you there are choices, not what the choices taste like.

Six names with one-line blurbs is a lot of labelling and not much guidance, and I'd rather have one sample reply per model than a slogan.

The Roleplay model reply is plainer and written in the first person. It also writes the user's lines for them, as in "I can cook up a mean feast!…". I dislike that habit in any app, because it is the chat version of a client who records your voiceover for you and then asks why it doesn't sound like you.

My advice is simple. Use Flow Mode while it is free, try a novel for the fun of it, and treat Normal chat as the fallback rather than the main event. Switch off the suggested replies if you'd rather write your own half of the scene.

The honest summary is that Joyland is a good writers' room and a poor long-term partner. Come for a scene and you will probably leave happy. Come for a relationship and you will spend a lot of the time reminding it who you are.

Characters were also inconsistent across sessions. The same pirate was a poet on Monday and a caveman on Wednesday. As an editor I'd call that a casting problem more than a memory problem, and it made it hard to care about any one character for long.

What the app remembers today

Short-term memory holds in the September 2026 check. Asked after 2 turns, the bot correctly recalls the user's name and the meat feast. Usopp even turns the memory check into a joke: "He forgot your name three seconds after hearing it and only got it right because Sanji just said it."

That is a good line and a low bar. Two turns is a goldfish test, not a relationship test, and the 2024 drift started ten times further in.

The memory you can buy

The Memory feature is paywalled. The Memory chip in chat opens the paywall.

The help centre spells out the ladder. On Free, memory works only within the current conversation. Standard adds short-term memory across sessions, and Premium adds long-term memory that holds information "across weeks or months".

Outside reports are less generous than the help centre. Reviewers report memory collapsing at 5-20 messages, even on Premium. Others report drift on long chats that is worse on the free model, better on Premium and never fully reliable.

What I make of it

I read those reports as confirmation of 2024 rather than news. My knight lost his horse at about the same point where reviewers say today's bots lose the plot. Two years is a long time in this business, and memory is the one area where the complaints sound unchanged.

My advice is to treat memory as a paid feature with an unproven warranty. Keep your own notes on long stories, the way an editor keeps a shot log for a client who changes their mind. And if a character matters to you, pin the few facts it must never forget, rather than trusting the model to hold them.

The free tier is the right place to find out whether a character is worth that effort. If the bot cannot hold a horse's name for a week, no plan upgrade will turn it into a companion.

Voice

Voice is thinner. No dedicated live voice-call feature was found, and the documented voice feature is a per-character custom voice pack, text to speech, trained in Toolkit. Secondary sources describe text-to-speech messages with about 10 selectable voice profiles per character, and one puts it bluntly: "Neither offers live two-way voice calls... not on these two [platforms]."

Text-to-speech messages are the ADR of companion apps. The words are right, but they never quite sit in the character's mouth.

Video

Video is a contradiction in the paperwork. No video-generation feature was found in the help centre or the Terms. Yet the plan table lists bot videos for the top plan, so something exists behind the paywall that nobody bothers to document.

My view is that media is a sideline here. If pictures are why you want a companion app, Joyland is the wrong shop, and I'd not pay for a plan on the strength of a feature the help centre does not describe.

A cap that moves with the day is the worst kind, because you can't plan around it. It felt like a diner that closes the kitchen whenever the cook gets bored, and it always seemed to get bored on the days I was hungriest.

Free in September 2026

Free chat with the Free Model seems unlimited today: no message cap is hit in about 6 messages. Six messages is a small sample, and I would not call it a promise.

There are some free image attempts on web and mobile as well.

The better surprise is the model menu. The model selector lets a free account choose "Roleplay" without paying, and the Roleplay model replies on the free tier. Flow Mode is labeled "Limited-Time Free".

Novels are free at least for the first chapters.

On paper that is a better deal than my 2024 cap, with one caveat. A credit budget and a message cap are counted differently, and a free plan that runs on credits feels generous right up to the evening you want a picture.

The price of all that is ads. The free tier shows interstitial ads, the full-screen kind rather than a banner you can scroll past. They are the tax on free, and I'd call it a steep one for anyone who reads slowly.

What free is good for

Free is good for exactly the way I used Joyland: short sessions, a scene or two, and no plans for a saga. In 2026 it looks more generous on paper than the free tier I had, with a better model and the best feature on the site thrown in for now.

It stops being good the moment you want the app to remember you, or to go a week without a full-screen ad. Those are the two things the paywall is built around, and I think Joyland knows it.

My advice is to spend the free weeks on Flow Mode and on a novel or two, because both carry a "for now" label. If they go behind the paywall later, you will at least know whether they were worth paying for.

The lesson I'd pass on is plain: find out what a paid plan changes before you pay for it. Joyland's upgrade changed the quantities and left the thing that annoyed me alone.

The prices on 23 September 2026

Prices are shown in USD in the upsell modal. Standard costs $9.99 a month, next to a struck-through $19.99. Premium costs $19.99 a month, next to a struck-through $39.99.

Standard annual costs $55 a year, shown as $0.15 a day, and the card says "Save $64.88/year". Premium annual costs $95 a year, shown as $0.26 a day. Its card says "Save $144.88/year" and "60% OFF".

The help centre, checked on 22 September 2026, tells a different story. It lists Standard monthly at $19.99 and Standard annual at $107.99, or $8.99 a month. It lists Premium monthly at $39.99 and Premium annual at $191.99, or $15.99 a month.

Reading the discount

So the help-centre prices are the struck-through "originals" in the modal. I see two readings. Either the discount is permanent, in which case the originals are decoration, or it is not, in which case the price after the countdown is anyone's guess.

The yearly savings have their own trick. Doing the math out loud, the "Save" figures are measured against twelve months of the already discounted monthly price, not against the help-centre price. That is a discount on a discount, which is how used-car lots do it.

What each plan adds

Both plans include "Intimate Themes Access" and both are ad-free. Both include "Suggested Replies Anytime" and "Regenerate Anytime". Both include access to Advanced Models, "Faster Response" and "Pro Models with Long Memory".

The differences are counts. Standard gets 5000 credits a month, where Premium says "No limits". Unlocking bot photos is "By credits" on Standard and "No limits" on Premium.

Pin messages go from 20 on Standard to 50 on Premium. Customized memory goes from 800 to 1500. Personas go from 20 to 50.

Read as a spec sheet, the gap between the plans is mostly volume. Standard already gets the ad-free screen, the better models and the content setting, and Premium doubles or more the counts on top. It is the difference between a mid-size and a large popcorn: same film, same seat.

The help centre's version

The help centre describes the same plans with other words. Standard and Premium include unlimited text conversations. Standard includes 20 image messages daily and 10 free custom packs monthly. Premium includes unlimited image messages and unlimited custom packs.

The two sources disagree on memory. The help centre gives Standard short-term memory only, while the plan table lists "Pro Models with Long Memory" for both plans. I'd trust neither until the chat itself proves it.

Then and now

In 2024 Standard left the filter exactly where it was. In 2026 the paid plans sell the content setting itself, as Intimate Themes Access, which is a strange reversal for an app that once refused my knight a sword fight.

Paying

Payment runs through an Antom checkout. The checkout merchant is "INFMIND PTE. LTD.". The pricing modal also has Apple Pay and Google Pay buttons.

Checkout reaches the card form showing "Joyland Standard Membership, Total $9.99". The check stops at the card form, and no payment is made.

A secondary source, geniusfirms.com, says a credit card is required and PayPal is not supported, adding that "the lack of PayPal support is a significant drawback". The Apple Pay and Google Pay buttons suggest the card part is out of date, but I'd not expect PayPal to appear.

The merchant name matters more than the payment method. A charge from INFMIND will not look like Joyland on a statement, so it is worth writing the name down before you pay, the same way you'd keep a receipt from a flea market stall that has no sign.

Is the step up worth it?

Standard at $9.99 a month is the only plan I'd consider, paid monthly, and only if images or the ad-free experience matter to you. Premium mostly buys bigger counts and a video feature nobody documents.

My 2024 month is the cautionary tale here. I paid for Standard hoping it would change how the app behaved, and it changed only how much of it I could have. Before paying, decide which single problem you want the money to fix, and check that the plan table actually names it.

I'd skip both yearly plans. The discount looks good until you read the refund clause, and a year is a long time to commit to an app whose memory I would not trust past a week.

After the 10 free chances on the Create-a-bot page, avatar generation costs credits. Text is fuzzier. One secondary source, techshali.com, estimates about 1 credit per text message, and nothing else confirms it.

Doing the math

At 30 Credits an image, Standard's 5000 monthly credits cover well over a hundred in-chat pictures, or ten permanent companion unlocks at 500 each. That is a fair amount for a plan at $9.99. I'd still spend them on unlocks for the one or two characters you actually use, since that purchase does not run out.

What you can't see

The UI shows no credit balance anywhere checked: settings, profile or the task page. No separate credit-pack store is found. So credits come from a plan or from tasks, and you are left to guess how many you have.

For an app that charges per picture, that is an odd thing to leave out.

It is the flea-market problem: prices are clear, but there is no sign telling you how much is left in your pocket. A currency with no visible balance and an expiry date is a gift card with the amount scratched off. You know it is worth something, and you find out how much only when it declines.

My advice: spend earned credits within the month, and prefer purchases that last. A permanent unlock is worth more to me than a stack of single pictures that I would have to count by hand.

It was a small moment and a clarifying one. A deadline that resets is not a deadline, it is a screensaver, and after that night I read every "limited time" label in the app as decoration.

The discount today

The upsell shown right after sign-up advertises 60% off. A new account gets a limited-time discount, a "Subscribe -60%" badge on the pricing modal. The pricing modal shows a countdown timer.

About 1 hour after signup, the timer reads about 50:30 in minutes and seconds. So in 2024 the timer promised 90% and in 2026 it promises 60%. The number changed and the theatre did not.

The percentages vary by line. Standard is shown as 50% off in the price lines, $9.99 against a struck-through $19.99. The Standard annual card says "55% OFF".

Premium is shown as 60% off, $19.99 against a struck-through $39.99.

No coupon code is offered, and there is no promo-code field. There is no welcome email offer either. I don't miss coupon codes here, because every price in the modal already arrives dressed as a bargain.

Free credits from tasks

A Task page sits under the user menu, at /billing?tab=task. The daily login reward is claimed automatically. The amount of the daily login reward is not shown.

"Share your story" gives +20 credits, daily. Watching a rewarded ad gives +15 credits per ad, and rewarded ads count up to 20 a day. That adds up to as much as 300 credits a day.

A rewarded ad plays in full and the counter moves to 1/20. The mechanism works, in other words, and it is not a fake button.

Referrals, Discord and creator milestones

Referral pays 50 credits for each invitee who registers via the referral link and logs in. The referral link is the site address with your UID added as a ref parameter. Joining the Discord and sending a code there gives 200 credits, one time.

Creators get milestones. Five users chatting with your bot gives +50 credits. Fifty users give +400 credits, and 100 users give +1000.

Are they worth it?

Watching all 20 ads a day earns 300 credits, which is ten in-chat images at 30 Credits each. It is a vending machine that pays you in coupons for the same vending machine. Twenty ads is also a lot of an evening to give a company that already charges for the ad-free version.

The one-time Discord bonus is the best return for the least effort, and I'd take it. The ads I'd only watch if they were already playing in the background of an evening with nothing better in it.

The referral bonus is fine if you have friends who would enjoy the app anyway. I would not recruit anyone for 50 credits, which is less than two pictures.

The countdown discount is real in the sense that the checkout charges it. It is not real in the sense that it runs out. I'd treat the discounted price as the actual price and decide on that basis, without the clock.

For an editor that is like getting a broadcast standards note on a cooking show because someone picked up a knife. The worst part was not knowing which word had set it off, and the uncertainty was worse than the rule.

What the settings allow in September 2026

Today the picture is reversed. Does Joyland AI allow NSFW content now? In practice, yes: the site is full of adults-only bots with graphic sexual content. There is no content toggle at all.

There is no age check. There is no adults-only confirmation anywhere, even for logged-out visitors.

"Intimate Themes" is paywalled. Premium's "Intimate Themes Access" is the paid lever for the adult side. Intimate Themes itself was left untried on 23 September 2026, so I can't say what the paid setting changes.

Bot text uses spellings built to slip past word filters, with a Greek letter or a digit swapped in. That tells you what the bot makers expect from the filter, which is not much.

What the policy says

The help centre encourages safe-for-work content and warns that adult content may reduce a bot's chances of being featured on the homepage. The research record classes the policy on adult content as allowed. The App Store content rating is 17+, and Google Play shows PEGI 18; those are store ratings, not a site policy.

The ToS list what is banned: - characters who are not adults, including "aged-up" ones - photorealistic sexual images and images of genitals - sex trafficking - gore and two extreme fetish categories - sexual content with animals, with a stated exception for "fictional species like Pokémon" - hate - promotion of self-harm

An Acceptable-Use clause bans asking anyone below legal age for personal information, and it bans depicting anyone below it. The ToS also ban posting real images of individuals or AI images that are too realistic, with no allowance for images of genitals.

The home page shows bully-themed bots whose backstory is set before the characters were grown up. It also shows step-family themes. A comment under one bot asks, in broken English, whether "age is just a number".

The ToS ban characters below legal age and "aged-up" characters, but the platform's own minimum registration age is 13. All such content is avoided during the September check.

My view

In 2024 a sword fight was too much for this filter. In 2026 the front page carries themes its own rules forbid, and the adult side is sold as a feature. I'd call that a filter that swung from paranoid to absent, with the paywall standing where the moderation should be.

What bothers me most is not the adult side itself. Adults choosing adult bots is their own business, and plenty of people come here for exactly that. It is the lack of any door between that material and a first-time visitor, on a site whose own registration floor sits well below adulthood.

The 2024 filter was annoying, but at least it was a decision someone had made. The 2026 feed looks like the absence of one. I'd rather have the jumpy filter back than a front page that nobody seems to curate.

I can't recommend Joyland to anyone who wants control over what they see. There is no switch to flip, so the only content setting that works is closing the tab.

The company named in the ToS differs from the checkout merchant, INFMIND. The App Store seller is ELDORIA PTE. LTD., with "(c) 2026 ELDORIA Inc" in the page footer, while the ToS and Google Play name MINDZEN, and no explanation is published.

Who may register

The ToS bar registration below 13, or below 16 for EU citizens and residents. There is no adults-only requirement in the ToS or in the Privacy Policy. For an app with this front page, that gap is the clause I'd fix first.

Money: refunds and free use

On billing, the ToS say "all fees are non-refundable". They add that refunds cannot be claimed for any part-used month or year of a subscription period. In plain English, a yearly plan is a bet with no exit, and a monthly plan is a smaller bet with the same rule.

The ToS describe "Free Use" as aimed at users who cannot afford any of the paid subscription programs. The ToS also say: "We reserve the right to deny free use to anyone at any time." I find that framing a bit sour: the free tier is the shop window, not a charity.

Your characters and your content

Users own their Characters and Generations. They also grant Joyland and other users a "perpetual, irrevocable" worldwide license to their content. Owning something you have licensed to everyone forever is like owning a car the whole street has keys to.

Your chats and your data

Dialogue text may be converted "into anonymous numbers and graphs" for academic cooperation with third-party institutions. Data is shared with vendors, with affiliates and in business transfers. Servers are in the United States.

The Privacy Policy says: "We do not and will not sell your Personal Information." Advertising networks use cookies. No fixed data retention period is stated in the ToS.

No clear statement that chats are used to train the AI was found; the academic-use clause is the closest. I'd assume the words you type leave the chat window in some form, anonymised or not.

The privacy promise that doesn't fit

The FAQ claims "no one — not even Joyland staff — can access your conversations". That claim does not fit the reporting and review language in the ToS, or the academic-use clause in the Privacy Policy.

Disputes

Disputes go to binding arbitration at SIAC, seated in Singapore. The ToS include a class-action waiver. They also set a 1-year limit for claims.

In plain English: a user with a billing complaint argues it in Singapore, alone, and within a year. That is a lot of paperwork to chase a monthly fee, which I suspect is the point.

None of these clauses is unusual on its own. Together, with the licence and the refund rule, they describe an app that keeps its options open and yours narrow, and I'd pay for it only in amounts I would not miss.

It sounds trivial until it happens with a code that expires, on a phone, with the email app in the other window. It is a small thing and a telling one. An app that interrupts its own sign-up to sell you something else has put its priorities in the order an ad network would choose.

In editing terms, it is a mid-roll ad dropped into the vows.

Upsells everywhere

The upsell modal appears on signup, on Toolkit, on the Joyland AI search page and on Memory. The upsell modals use a fake-urgency countdown. The struck-through "original" prices look inflated.

Four doors into the same shop would be fine if the shop were honest about its prices. With a countdown that resets and originals that look invented, every modal feels like a car salesman who keeps checking with his manager.

The Memory modal annoys me most. A user who taps Memory is asking why the character forgot something, and the answer on offer is a countdown and a price. That is not help, it is a sales pitch at the exact moment the product let you down.

The Google Play listing is linked from the footer too. Support runs through [email protected]. A 5th footer icon, in light blue, opens nothing when clicked.

A dead icon is a trivial bug. It is also the kind of thing a client spots in the first viewing, and it tells me nobody has clicked through the footer in a while.

Apps, names and languages

Joyland also has an Android app on Google Play, com.joyland.ai. The Terms of Service also mention an iOS app. The iOS app "Joyland AI: Anime roleplays" is on the App Store.

The app is also listed as "Joyland AI: Anime roleplays" and "Joyland: Chat with AI Character". Two product names sit on top of three company names, which makes searching for help or a charge harder than it needs to be.

The iOS listing names English, French, Japanese, Korean, Simplified Chinese and Spanish as languages. No language switcher was found on the Joyland AI official site itself. No developer or API documentation was found either.

My take

None of these would make me leave on its own. Each is the kind of note a client leaves on a second cut: fixable, a bit embarrassing, and a sign of where the attention went. Together they add up to an app that tidies the storefront and neglects the back office, and I'd budget some patience for anything involving support or billing.

The practical upshot is to keep the support address and the payment partner's name somewhere handy before you pay. If something goes wrong with a charge, you will want both, and the footer will not do the finding for you.

It is not for anyone who wants an ad-free or privacy-minded experience on the free tier. It is not for anyone who wants content controls or a safe-for-work feed, since adults-only bots sit on the front page and there is no toggle.

It is not for anyone who is not an adult, because the site has no age gate despite its adult content. And it is not for people who want refunds or easy cancellation: cancelling runs only through the payment partner, and there are no refunds.

The last two points are the ones I'd weigh hardest. Ads and content can be avoided with money or care; the terms on refunds and cancelling apply to everyone who pays.

If you are in the first group, the free tier will tell you in a week whether Joyland fits. If you are in the second, no plan fixes it, and I'd not waste the week. Better to spend it on an app built for what you actually want.

There was no argument with a support form and no hunt for a hidden button. The app had become a snack I no longer reached for, and the timer kept its countdown going for an audience of none, which is the most Joyland ending I can think of.

How renewal and cancelling work today

Subscriptions auto-renew "automatically until canceled". A subscription can be cancelled only through the payment partner it was bought from. It must be cancelled before the next billing period to avoid another charge.

A secondary lead older than 60 days, aigirlfriendscout.com from February 2026, gives in-app steps: Settings, then Manage Subscription, then Unsubscribe, then Confirm. That clashes with the ToS, which send you to the payment partner. I'd try the Settings entry first and then check with the partner that the renewal is really off.

A secondary lead older than 60 days also says the bank statement shows "Joyland Ai"; that is unconfirmed. Given the checkout merchant name, I'd look for INFMIND on a statement as well.

Account deletion sits under Settings, "Remove account". Cancel the subscription first, since deleting the account is not the same as stopping the payment.

Before you pay

• Note which door you paid through: the web checkout, Apple Pay or Google Pay.
• Pay monthly, because fees are non-refundable and a year is a long bet.
• Set a reminder a few days before renewal, and cancel with the partner, not just in the app.
• Take a screenshot of the cancellation screen, since the ToS leave the details to the partner.

None of that is exotic. It is the same routine I'd suggest for a gym membership, and it matters more here because the terms offer no refund if the reminder comes a day late.

Use personas instead of new characters. Being a different "you" with the same bot gives more variety than hunting for a new bot, and the Persona chip costs nothing.

Keep your own notes for anything longer than an evening. Memory drifts, so a short recap you paste back in beats hoping the model remembers.

Switch new bots to private while you build them, and turn off the leaderboard and follower toggles unless you want an audience.

Settings has a dark-mode toggle. It is the one setting I'd turn on before anything else, because the ten minutes before bed is when this app works best.

Take the one-time Discord credits, skip the ad marathon, and spend earned credits within the month before they expire.

Try a novel before paying for anything. The first chapters are free, and a story with choices shows the writing at its best without asking the model to remember much.

And read a public bot's Story feed entries before a long chat with it. Someone else's session is the cheapest preview you will get, and it costs nothing but a few minutes.

Memory is weak. It holds a few turns and fades over a longer chat, and the better memory is sold rather than shown.

The free tier is good by the standards of this category. It includes a real model choice and the best feature for now, at the cost of full-screen ads and a credit system with no visible balance.

Value for money is mixed. The discounted monthly price is fair for what Standard adds, while the yearly plans and the struck-through originals are not.

Images and voice are weak. Pictures depend on paid packs, voice is text to speech, and video is listed but not documented.

Privacy and terms are weak: a broad licence over your content, vague privacy promises and three company names. Cancelling and billing are weak too, with no refunds and cancelling only through the payment partner. Content settings are weak because there are none to set.

Rewarded-ad credits, up to 300 a day, are unusual too. Paying people to watch ads is a very Joyland solution: generous on paper, tedious in practice, and a clear sign of where the money really comes from.

Where Joyland is weaker is the part the table below cannot show: keeping one character consistent for weeks. That is the job companion apps are sold on, and it is the job Joyland does least well. It puts its work into the shelf instead, and the shelf is enormous.

It is also weak on the paperwork. Three company names, a front page with no age gate and no switch, and a refund clause that says no: I'd want better from any app before handing it a card number.

So I'd pick Joyland for a quick scene or a bit of fan-fiction fun, and something else for a character worth getting to know over a month. That is the same split I would have drawn in 2024, only with better tools on Joyland's side of it.

If you already pay for a companion app you like, Joyland is worth a free week as a side dish rather than a replacement. It is the diner you go to for the pie, not the one you go to for dinner.

More on this site: the best AI girlfriend apps, free AI chat apps, the best AI companion apps, AI roleplay chat apps, all AI companion app reviews.

My lasting thought is that an app built on user-made characters is only as good as the shelf it keeps. Joyland has the best shelf-browsing in the category and the weakest shelf-keeping, and those two facts are the same fact seen from different ends.

That is not a reason to avoid it. It is a reason to use it the way it works best: lightly, briefly, and with your eyes open.

I'd still open it for ten minutes before bed. I'd just keep the card in the drawer.