Candy AI pre-payment risk check · assessed 2026-07-19
Candy AI has current public product, subscription, and browser/PWA help pages, so there is a documented service footprint. That is not a safety, privacy, operator, scam, or reliability verdict. Companion Curator has not tested payments, moderation, account outcomes, or long-term use; privacy, export, refunds, deletion, content-rule enforcement, and all-in token cost remain unresolved.
The useful answer is not one badge. A visible service, predictable billing, clear content rules, acceptable data terms, and a workable exit are separate questions. Evidence for one does not fill the others.
The five-question answer
“Documented” means a current source supports that narrow statement. It does not mean Companion Curator reproduced a purchase, tested a workflow, verified the operator, or measured a typical outcome.
On a narrow screen, swipe the table sideways to read all four columns.
| Question | Current record | Does not prove | Your next check |
|---|---|---|---|
| Is there a publicly documented Candy AI service? | Documented footprintYes—current product, subscription, and help pages document browser access plus an installable home-screen web app. | Operator identity, legal compliance, a successful purchase, reliable delivery, or a not-a-scam verdict. | Confirm the exact candy.ai hostname, seller, charge descriptor, and terms before entering payment details. |
| Is the cost predictable before paying? | Base charges onlyFive web charges are recorded, from weekly through annual, and each lists 100 generic monthly tokens plus unlimited text. | What a token buys, top-up prices, failed-output treatment, tax, renewal totals, or typical all-in spend. | Capture the checkout total and model one realistic media task, including retries, before choosing a longer term. |
| Are content and moderation boundaries established? | Boundary incompleteThe current product record documents adult-first romance, roleplay, adult roleplay, and media features. | A rule-free service, moderation effectiveness, response time, consistent enforcement, or safety for a particular use. | Find and read the current governing content and reporting rules for your lawful, consensual use before sharing material. |
| Is privacy or export established? | UnknownNo. Privacy-policy and portable-export status are unresolved in the current canonical account record. | That the service has no policy, is private, is unsafe, uses encryption, or deletes every retained copy. | Read the current privacy, retention, training, disclosure, and export terms before sharing sensitive information. |
| Can a buyer cancel, request a refund, or delete the account? | Partial procedureThe publisher documents settings-based cancellation. Refund, export, and account-deletion paths remain unresolved here. | That renewal stops, a refund is granted, an account is erased, or retained data is deleted. | Locate cancellation and every other exit control before paying; treat cancellation, deletion, refund, and erasure as separate tasks. |
The “app” question changes the risk
Candy's current help record says iPhone and Android users can install its browser experience to the home screen. A home-screen icon does not establish a native package, store oversight, identical features, notifications, billing, or policy behavior.
Desktop / web
Availability: available.
Installed implementation: not applicable.
Broader feature, content-policy, billing, notification, and account-control parity remains untested.
iPhone
Availability: available.
Installed implementation: not applicable.
Broader feature, content-policy, billing, notification, and account-control parity remains untested.
Android
Availability: available.
Installed implementation: not applicable.
Broader feature, content-policy, billing, notification, and account-control parity remains untested.
Five charges, one unresolved usage model
Candy currently lists weekly, two-week, monthly, quarterly, and annual web commitments. The quarterly and annual amounts are labeled promotional listings, so neither is treated here as an established renewal total. Each Premium row says unlimited text and 100 generic tokens per month, but the source does not say how many images, voice messages, videos, retries, or failed outputs those tokens buy.
Model Candy AI costCash charge by commitment
All five are observed web USD listings checked 2026-07-13. Tax, promotional eligibility, discounts, renewal presentation, checkout outcome, token conversion, and typical total spend remain unresolved.
The largest unresolved risks
The current product record supports romance, roleplay, adult roleplay plus text, memory, generated media, voice messages, video, and persona controls.
Feature availability does not establish quality, lawful use, consent handling, moderation, or consistency.
Current product pages document adult-oriented modes and media features. The canonical Candy record does not yet contain a current moderation, prohibited-content, age-eligibility, or age-verification policy source, so this guide cannot claim who may use it, what requests are prevented, how reports are handled, or whether rules are enforced consistently.
The canonical record does not currently establish a privacy-policy source or general export path. That gap supports neither a privacy guarantee nor a conclusion that harm occurred.
Settings-based cancellation is documented. Refunds, data export, account deletion, retained-data treatment, and successful outcomes remain unresolved.
Public reports add questions—not a verdict
The current sample contains 3 direct Reddit items and 3 YouTube creator items. Its direction mix is 1 favorable, 1 critical, and 4 mixed—but a purposeful sample cannot be converted into a satisfaction rate.
The current sample contains six deliberately selected Reddit and YouTube items. Every item has a commercial, methodological, or product-identity concern, so the sample is a risk checklist—not representative sentiment, a scam check, or a safety verdict.
Read the complete Candy AI report synthesisSelected reports describe filtering, feature changes, and privacy tradeoffs, while creator summaries sometimes compress policy details.
Change-sensitive; official policy controls factual claims.Token-funded media and paid scenes repeatedly change perceived value beyond the base subscription.
The friction recurs; typical burn and total monthly spend remain unknown.Affiliate links, referrals, opaque scoring, reused channels, methodology gaps, and similarly named domains reduce independence.
High confidence that source hygiene is necessary; no coordination claim is made.One creator reports an easy cancellation, but no workflow artifact or independent outcome is available.
Too little evidence for an easy-or-hard account-operation verdict.A seven-step preflight
Short answers, same evidence ceiling
Current sources do not support a binary safety verdict. They document a browser/PWA service, prices, features, and cancellation instructions, while privacy, moderation outcomes, deletion, refunds, security, and long-term use remain untested or unresolved.
The record documents a public service footprint but does not independently verify the operator, purchases, delivery, refunds, charge disputes, or long-term outcomes. That is not enough for either a scam or not-a-scam verdict.
Candy documents browser access and an installable home-screen web app on iPhone and Android, and says it does not offer a native App Store or Google Play app. A home-screen icon does not prove native implementation or feature, billing, notification, or policy parity.
Current web listings span weekly, two-week, monthly, quarterly, and annual charges. Each lists unlimited text and 100 generic tokens per month, but the source does not establish token conversion, top-ups, failed-output treatment, tax, or typical all-in spend.
The publisher documents settings-based cancellation. The current canonical record does not establish refund, export, or account-deletion procedures, and Companion Curator has not tested any outcome.
Exact evidence ceiling
Official pages support only the narrow product, delivery, price, and cancellation facts above. Selected public reports support only the disclosed qualitative sample. Neither source class can manufacture a safety, operator, privacy, moderation, or account-outcome verdict.
This exact source set first requires refresh on . Removed Trustpilot rows, banned-subreddit items, search snippets, cached fragments, affiliate terms, and similarly named store products are excluded.
Continue the decision