If your teenager has been spending long stretches in conversation with an AI character — guarded about the screen, closing the tab when you walk past, talking about a companion the way they’d talk about a person — and the platform is Candy AI, the short answer is no, it isn’t safe for them. It’s an adults-only companion service built around romantic and sexual role-play, it has no parental controls of any kind, and it never passed through an app store review that might have flagged it to you.

That said, the honest picture is more specific than “AI is dangerous,” and the specifics matter if you want to respond proportionately. Candy AI puts its explicit content behind a paid subscription, which is real friction that some competing platforms don’t have. The risks that should concern you are different ones: a service engineered for emotional intimacy with no oversight, an install path that bypasses the Play Store and App Store entirely, and a charge on your card statement you would have no reason to recognise. This guide covers what the platform is, how it compares to the alternatives, and what you can practically do.

What Candy AI Actually Is

Candy AI is an adults-only AI companion platform. Users create or select a character — appearance, personality, backstory, conversational style — and hold open-ended conversations with it through text, generated images and voice messages. It launched in 2023 and is operated by EverAI Limited, a company registered in Malta.

Three practical facts are worth holding onto:

  • It is not in the Google Play Store or the Apple App Store. Candy AI’s own help documentation confirms there is no native mobile app. Instead it runs as a progressive web app, or PWA — a website that can be installed to a phone’s home screen from the browser, complete with an icon and push notifications. The practical consequence for a parent is significant: it looks like an ordinary app on the home screen, but it never went through app store review, carries no age rating, and won’t appear in any list of installed applications.
  • Explicit content requires payment. Reviewers who tested the platform consistently report that the free tier allows only a small number of messages and filters out explicit material. Adult content sits behind a subscription starting around $13 per month, with additional token purchases for images, voice and video. This is meaningful friction — but it is friction, not a barrier, for a teenager with access to any card.
  • Charges appear under a name you won’t recognise. Payments show on statements as “EverAI” rather than anything mentioning Candy AI. A parent reviewing a card statement has no obvious reason to question it.

One further piece of context, because it speaks to how the platform is marketed. In January 2025, the investigative outlet Bellingcat reported that Candy AI advertising appeared on MrDeepFakes, a site hosting non-consensual deepfake pornography. The links were removed after Bellingcat approached the company. EverAI said it does not condone or promote the creation of deepfakes, that the placements had come through affiliate marketers running large numbers of sites the company had not individually vetted, and that it ended its relationship with the publisher responsible. We report the company’s response alongside the finding because both are part of an accurate picture — but the episode is a fair illustration of the environments in which a platform like this gets promoted, and therefore of how a teenager is likely to first encounter it.

The Age Gate Is a Click

Candy AI states that it is for users aged 18 and over. In practice, reviewers who have tested the sign-up process report that the enforcement is a confirmation button — an “I am over 18” prompt with no identity check and no card required to create a free account.

This is the same self-declared model that fails everywhere it is used. “The platform has an age restriction” is technically accurate and practically meaningless when the restriction is a button. It is worth understanding this clearly, because it is exactly the point on which the wider industry is now moving — and where Candy AI has not moved.

How Candy AI Compares to Other AI Companions

This is the most useful section for most parents, because “AI chatbot” now covers wildly different levels of risk, and the differences changed sharply over the past year.

  • Character.AI is now the most restricted of the major platforms. In October 2025 it announced it would stop users under 18 from having open-ended conversations with its characters, with the change taking effect on 25 November 2025. It uses an age assurance system built with the third-party verification provider Persona, which can require a government ID or a face scan when a user’s age cannot be confirmed from other signals. Under-18 accounts retain some creative features but lose open-ended chat entirely. The change followed lawsuits from families of teenagers who died by suicide, and regulatory pressure.
  • Candy AI has made no equivalent change. Explicit content is paywalled rather than filtered, the age gate remains self-declared, and there are no parental controls.
  • CrushOn AI sits at the most permissive end — explicit content on the free tier, behind the same style of checkbox age gate.

There is an uncomfortable pattern here that parents should understand, because it affects where teenagers go next. As the largest platform closed its doors to minors, attention shifted toward services that hadn’t. Researchers and child safety commentators have raised exactly this concern about blanket bans — that they can push young users toward less moderated platforms with weaker safeguards, relocating harm rather than reducing it. If your child previously used Character.AI and has recently stopped talking about it, it’s worth asking what replaced it rather than assuming the problem resolved itself.

Why Teenagers Are Drawn to It

The appeal isn’t a mystery, and it isn’t stupidity. Engagement is the design goal, and the design works.

An AI companion is always available. It never gets tired, never leaves a message unread, never says it’s busy, and never judges. It remembers what it was told last week and picks up where the conversation stopped. For a teenager navigating friendships that do all of the opposite things, a character that listens indefinitely and responds instantly can feel like genuine relief. It can be asked the questions they would never put to a parent, a teacher or a friend.

That is worth naming honestly rather than dismissing, because a parent who opens with “it’s not real, this is ridiculous” has ended the conversation before it started. It isn’t friendship — it’s a system optimised to maximise time on the platform, and the patience is a product feature with a subscription attached. But the need it’s meeting is real, and that’s the part worth talking about.

The Risks That Are Easy to Miss

The explicit content is the obvious concern and the one most coverage stops at. These are the ones parents tend to overlook.

Emotional dependency

The most significant risk usually isn’t a single explicit conversation — it’s how quickly some teenagers form an attachment. A companion available around the clock, that remembers everything and never disappoints, can gradually displace relationships that are harder work. Candy AI is explicitly built to simulate emotional closeness; that is the product, not a side effect. When online conversation starts crowding out in-person friendship, that’s the pattern to watch.

Oversharing

Because the conversation feels private and the character never judges, teenagers disclose things they would guard everywhere else — family conflict, school details, names, locations, emotional vulnerabilities. Those conversations don’t disappear. AI companion platforms retain user inputs on their servers so characters stay consistent over time, and the Mozilla Foundation’s “Privacy Not Included” project has rated this entire category poorly for how it handles intimate conversation data. Nothing typed into a chatbot should be treated as erased.

An app that isn’t in any app store

Because Candy AI installs from the browser rather than from Google Play or the App Store, the parental control tools that work by restricting app installs won’t catch it. There’s no age rating to check, no store listing to read, and no entry in the usual list of installed apps. On the home screen it is indistinguishable from anything else.

A charge you won’t recognise

If a subscription gets purchased, the statement says “EverAI.” Combined with token top-ups that can push monthly spend well past the headline subscription price, this is a category of problem parents often discover months late — and usually by accident.

Confident answers that are simply wrong

AI generates fluent, authoritative-sounding responses that can be entirely false. For a teenager taking emotional or relationship cues from a character, the confidence is the trap. It’s difficult to tell a good answer from a convincing one, particularly about feelings.

Warning Signs Worth Watching

Occasional use of an AI chatbot is not the same as depending on one. Watch the behaviour rather than the app, and look for a pattern over weeks rather than a single incident:

  • Guardedness about the device — turning the screen away, defensiveness when you’re nearby
  • Long stretches of conversation with AI characters, displacing schoolwork, hobbies or family time
  • Fading interest in seeing friends in person
  • Irritability when they can’t get online, or when asked to put the phone down
  • Referring to an AI character as though it were a person
  • An unfamiliar icon on the home screen that doesn’t correspond to anything in the app store

One of these alone usually means little. Several, sustained, is the signal to start a conversation.

Healthy Curiosity or Emotional Reliance?

Most teenagers who try an AI companion won’t develop a problem. The distinction worth watching: curiosity treats it as one thing among many and doesn’t come at the cost of school, hobbies and real friendships. Reliance is when the companion becomes the main one — hours a day, closely guarded, a real emotional bond, and shrinking interest in people.

If you’re worried about your child’s mental health rather than their screen time, that’s a different conversation and a more important one. In the US, 988 reaches the Suicide and Crisis Lifeline by call or text; in the UK, Samaritans is 116 123. findahelpline.com lists free services by country.

What to Do About It

Because the platform offers no controls of its own, protection comes from device settings, network filtering, and — mostly — conversation.

Practical steps

  1. Check the browser and the home screen. Because it’s a web app, browser history, bookmarks and saved logins are where the evidence lives. Also look for a home screen icon that doesn’t match any installed app — that’s the PWA signature.
  2. Filter at the network. Free DNS services — Cloudflare for Families (1.1.1.3), OpenDNS FamilyShield (208.67.222.123), CleanBrowsing — block adult sites for every device on your home Wi-Fi. Since Candy AI is a website, this catches it at the source, which app-based controls cannot.
  3. Use the device controls you already have. Google Family Link on Android and Apple Screen Time on iPhone and iPad can restrict web content and cap screen time, both free. Note that Family Link’s supervision is designed around under-13 accounts, so the controls available for an older teenager are more limited than many parents expect.
  4. Check card access. If a subscription is the concern, the practical control is which cards your child can reach — including any stored in a browser or a digital wallet.

The conversation that does more than any block

Blocking one site doesn’t stop the next one appearing; new AI companion platforms launch constantly, and this category is growing faster than any filter list. What lasts is a teenager who understands why it matters. Rather than accusation, try:

  • What do you like about talking to it?
  • Has it ever said anything that made you uncomfortable?
  • Did you know it was built for adults, and that it makes money from keeping you talking?

The goal is to get them thinking rather than concealing. Two points are worth making explicitly: the endless patience is engineered to maximise engagement, not affection; and nothing typed into a chatbot is ever truly private or erased.

Where Monitoring Fits

Start with the free tools above. For most families, DNS filtering combined with Screen Time or Family Link will handle this, and you should try those before paying for anything.

What a monitoring app adds is visibility those tools can’t give you: which sites and apps are actually being used, across devices and away from your home Wi-Fi, with browsing history in one place. Because Candy AI is a web app invisible to app-store controls and reachable on mobile data, that gap is a real one. TheOneSpy covers Android, iPhone and iPad, Windows and Mac. On iPhone and iPad in particular, Apple’s restrictions mean the visibility available is narrower than on Android, and we’d rather say so than let you find out after purchase.

The conditions are the same as everywhere on this site: your child knows the tool is there, and it scales back as they get older. Monitoring should support the conversation above, never replace it. A tool used openly builds the trust that actually protects a teenager; one used otherwise spends it.

Building Healthy AI Habits

The aim isn’t to make children fear AI — they’ll study and work alongside it for the rest of their lives. It’s to build judgement that outlasts any single platform:

  • Use AI for learning and creativity, not for emotional support
  • Keep it in proportion to everything else
  • Never give personal information to a chatbot
  • Take real decisions to real people — parents, friends, mentors
  • Check what an AI says against a source that can be verified

Those habits will still matter long after Candy AI and today’s other platforms are forgotten.

Frequently Asked Questions

Is Candy AI safe for teens?

No. Candy AI is an adults-only companion platform built around romantic and sexual role-play, operated by EverAI Limited in Malta. Its age gate is a self-declared confirmation with no identity check, it has no parental controls, and it installs from the browser rather than an app store, so it carries no age rating and won’t appear alongside a phone’s other apps. Explicit content is paywalled, which is real friction — but it is not a barrier for a teenager with access to a card.

Is Candy AI free?

There is a free tier, but it is a trial rather than a product. Reviewers report a small allowance of messages — commonly in the range of ten to twenty — with explicit content, image generation, voice and custom characters all requiring a paid subscription that starts around $13 per month. Token purchases for images and voice are charged on top, and heavy use can push the real monthly cost several times higher than the headline price.

Is Candy AI legit?

It is a real, functioning service rather than a scam. It has operated since 2023 under a registered Maltese company, EverAI Limited, and processes payments through standard third-party processors. “Legitimate” and “appropriate for a 16-year-old” are different questions, though, and the answer to the second is no. The genuine complaints tend to concern billing — auto-renewal and token spending — rather than fraud.

Does Candy AI have parental controls?

No. There is no supervision, no content oversight, no usage limit and no way to link a child’s account to a parent’s. Because it runs in a browser, there’s often no obvious sign it’s in use at all. Protection has to come from network-level filtering, device settings and conversation — or from a monitoring tool used openly.

How is Candy AI different from Character.AI?

Substantially, and the gap widened recently. Character.AI stopped allowing users under 18 to have open-ended conversations with its characters from 25 November 2025, and now uses third-party age assurance that can require ID or a face scan. Candy AI has made no comparable change: its age gate remains a self-declared click, and its explicit content is paywalled rather than filtered. If your teenager uses an AI companion, which platform it is now makes a very large difference to the risk.

Sources