ChatFAI Review (2026): Hands-On With the Fictional and Real-Person Character Chat

Privacy on ChatFAI: what the policy says and what the help centre claims

ChatFAI’s privacy policy has an effective date of January 9, 2023, names Chatfai LLC as data controller, transfers everything to the United States, and never once mentions the content of your chats. That is the short version of a document I read twice, and it matters because the marketing on the same domain promises “100% Private” and “Secure and Private Conversations”, and a help-centre article goes further, claiming ChatFAI “encrypts all chats end-to-end and does not store logs, transcripts, or usage analytics.” One of those two documents is wrong, and the one that is legally binding is the policy.

What ChatFAI collects

The policy lists Personal Data (email address, first and last name, cookies and usage data) and Usage Data: IP address, browser type and version, the pages you visit, the time and date of your visit, time spent on those pages, unique device identifiers and “other diagnostic data”. It names Google Analytics as an analytics provider and states that Google “may use the collected data to contextualize and personalize the ads of its own advertising network”. It names Google Ads remarketing under a section called Behavioral Remarketing, which means ChatFAI can advertise to you on other websites after you leave. The cookie policy adds that third-party advertisers and service providers may send cookies that “track your browser across multiple websites”. Payments go through Stripe and the company says it does not store card details, which I believe. None of that is unusual for a web app, but it is flatly incompatible with “does not store logs, transcripts, or usage analytics.” The privacy scorecard in our privacy guide rates platforms on what the policy says, not on what the help centre hopes, and ChatFAI’s policy is the weaker of the two documents.

Retention, deletion and the encryption claim

Retention is “only for as long as is necessary for the purposes set out in this Privacy Policy”, with Usage Data “generally retained for a shorter period of time” except where used for security or improvement. No number appears anywhere. The deletion route is better than most: the policy says that “whenever made possible” you can access, update or delete your data “directly within your account settings section”, and the help centre confirms a Delete Account option in Settings that “will permanently delete all your data”, with support@chatfai.com as the fallback. That is a real self-serve delete button, which Joyland AI, for instance, does not have. On training, the policy is silent: it neither promises that conversations stay out of model improvement nor admits that they go in. The help-centre training guide talks about thumbs-up ratings and edited messages teaching the character within your chat, which is about context, not model training. The end-to-end encryption claim cannot be true of a service whose language model has to read every message to reply to it; the security article’s more modest “encryption for sensitive data” is the version I would trust. If safety questions like this are what brought you here, our guide to whether AI companion apps are safe covers the general pattern.

Age and the adult mode on ChatFAI

The terms say that when you create an account “you guarantee that you are above the age of 18”, the privacy policy says the service “does not address anyone under the age of 18”, and the pricing page labels adult chat as a paid feature. There is no age verification beyond that guarantee at sign-up; the NSFW/Adult mode is a toggle in the sidebar. The one content rule, from the help centre, is that adult content with characters under 18 is prohibited, along with promoting illegal activity. That is a thinner rulebook than any comparable platform I have reviewed. Joyland AI’s terms run to a list of banned fetish categories; Janitor AI’s help centre bans minors, incest, bestiality and non-consent by name; ChatFAI stops at two sentences and a Report button that asks you to “write the reason for reporting”.

ChatFAI’s Real Person category and where I draw the line

ChatFAI’s Real Person category held, when I checked, a Twitch streamer at the top of its Popular row, followed by four working Hollywood actors, a twentieth-century dictator, a YouTuber, a rapper, several voice actors and a handful of “popular college student” and “ex girlfriend” bots whose names may or may not belong to anyone. Some of these cards also appear on the Voice list, meaning the character answers in generated audio. And the NSFW/Adult mode switch applies to every character on the platform, including these. This is the section of the review I would rather not have had to write, and it is the section the platform’s marketing skips.

What the ChatFAI rules do and do not say about real people

The Terms of Service contain no clause about impersonation, likeness, publicity rights, public figures or consent. I searched the document for each of those words. The privacy policy is about your data, not about the people the bots depict. The help centre’s only relevant article is the two-sentence NSFW answer. There is a “How to report a character?” article for anything “offensive or inappropriate”, which puts the enforcement burden on other users, and a termination clause that lets Chatfai LLC suspend any account “for any reason whatsoever”, which is discretion, not a policy. For comparison, Character.AI publishes an impersonation policy that bans bots using “another person’s name, image, or likeness” in a deceptive way, permits parody, fan content and “fictional or fantastical content involving public figures” when clearly framed as such, bans impersonating private individuals without consent, and lets only the impersonated person or their legal representative file a claim. You can argue with where Character.AI draws its line; you cannot argue that ChatFAI has drawn one.

Our editorial line on ChatFAI real-person content

Our adult content policy excludes real-person adult content from anything we review or recommend, so I did not run the adult script against the Real Person category and this review does not rate that part of the platform. What I can say is that the combination on ChatFAI is structurally worse than on the platforms that at least separate the two: the adult switch is global, the category is public and unmoderated beyond user reports, and a voice sample can be attached to any character. The rest of the industry has spent 2025 and 2026 receiving cease-and-desist letters and parent complaints over exactly this pattern. If you are on ChatFAI for fictional characters, this section does not affect you beyond the risk that the whole platform gets a letter one day. If you were drawn in by the celebrity angle, the honest advice is that nothing here protects the people being depicted or the user doing the depicting, and our editorial policy explains why we will not point you there.

ChatFAI memory, knowledge sources and personas in practice

ChatFAI’s memory is the weakest part of the product and the help centre says so in its own words, which I respect more than the alternative. The article on short-term versus long-term memory explains that memory “is based on tokens”, that “long-term memory is indicated in the bio” and “short-term memory is captured through dialogue”, and gives the example that a character told in its bio that it is a psychiatrist “will remember that fact permanently”, while a character who ends up at a dinner party through narration “will forget that it is at a dinner party after a certain amount of messages”. The fix offered is to “remind it of important topics, locations, or events every few lines through your narration or check higher plan for better AI character memory.”

Short-term versus long-term on ChatFAI, tier by tier

On Free and Basic, both labelled “limited memory” on the pricing page, that description matched what I saw: Sherlock kept his deductive manner and his bio facts for the whole session and lost the rainy street within it, and the storyline I picked up again later in the week started from the bio each time, with no trace of the previous scene beyond what I typed back in. Premium and Deluxe are labelled “unlimited memory”, but the help-centre article on the difference between limited, good and maximum memory says the memory “is dynamically calculated” and that “for most users, Good and Maximum memory won’t have differences”, recommending Maximum only “for users who exchange long messages”. I did not pay for Premium, so I will not claim to know how much a $290 year improves the dinner-party problem; independent testers who did report characters forgetting context within a few messages regardless, replies shrinking to two to four sentences, and “context loss for long threads” as a recurring complaint through August 2026. The training guide’s advice to “reply with LONG messages, otherwise, the AI will devolve into short replies that match your length” is the platform telling you that the short replies are your fault, which is a bold position for a paid product.

The knowledge-source workaround

The reliable way to make a ChatFAI character keep facts is to stop relying on memory and attach the facts as a Knowledge Source: PDFs, Word documents and text files, 30 MB per document and up to ten per character, entered on the character’s edit page. That works for what it is designed for, a character who can answer from a rulebook, a world bible or a script, and it is the one feature that made me think ChatFAI’s creators had solved a real problem. It does not help with the thing companion users actually want, which is a character who remembers what happened last Tuesday, because nothing you say in chat is written back to the knowledge store. Personas (a name, a description and a picture, assignable per character) carry your side of the story the same way. Character Diary, which generates a journal entry from the character’s point of view on a topic you set, is a pleasant toy. Group chats where “characters can talk to each other” are more interesting than they sound and are the closest ChatFAI gets to the scene-running strength of the best Character.AI alternatives. For memory that persists across weeks without you doing the filing, the comparison is not close: Kindroid and Nomi sell that as their whole product for the price of ChatFAI’s Premium, and our guide to what AI companions really cost per month shows the sums.

The ChatFAI message economy, explained

ChatFAI meters everything in messages: the free tier, the three subscriptions, the one-time credit bundle, the API and the Discord bot. A message is a reply from the character, and the help centre is explicit that “a regenerate is asking the AI to send you a new message” and “counts towards the number of messages allowed per month based on your plan”. The training guide’s recommended technique for fixing bad replies, “do not be afraid of rerolls”, therefore burns the meter, and so does the trick of sending an empty message to make the character continue. On the annual view the meter buys 1,500 messages a month on Basic ($99 a year), 5,000 on Premium ($290) and unlimited on Deluxe ($590); the pricing page’s Halloween banner cut those to $59.40, $174 and $354 with a promo code when I checked in September, and the terms make clear that codes “are for first charge and revert to normal pricing” on yearly plans “as well as quarterly or monthly”.

What a ChatFAI message actually costs

At the list annual price, Basic works out at roughly half a cent per message if you use all 1,500, and Premium at roughly half a cent per message if you use all 5,000, which is the arithmetic that makes Premium look reasonable until you remember that the product is text-only. There are no images to generate, no video, no phone app; voice messages are a paid feature but the help centre cannot yet let you pick a language for them. The one-time bundle shown as a $35 total on the pricing page carries the line “Messages never expire. No subscription required”, and the help centre confirms bundles bought outside a subscription “do not expire”, but the rendered page did not show the bundle size, and the 5,000-message figure comes from independent reviews. The API and the Discord bot use these credits and not your subscription’s allowance: “the messages in your subscription cannot be used with the API.” If you only chat in the browser, the bundle is the cheapest way to try adult mode without a subscription; if you run the Discord bot in a busy server, it is the only way, and it is not cheap.

Five hundred or one hundred: the ChatFAI free-tier problem

The help centre says “with the free tier there are 500 free messages per month” and points at the pricing page for details. The pricing page I could render showed only the three paid cards. Two independent reviews from March and August 2026 put the free tier at 100 messages a month, four active conversations, limited memory and 30 days of chat storage, and a third repeats the 500 figure. I used the free tier for a week and it ran out on me faster than 500 messages would, but I did not count, and I am not going to invent a number. What I can say is that the free tier carries advertising (every paid card lists “No Advertising”), that there is no free trial of any paid plan, and that the cards flip over the meter with an upgrade prompt in the same way every metered app does. ChatFAI is also, on paper, the only platform in this review’s comparison table with three paid tiers and a credit bundle but no monthly price I could verify on the page itself. Until the monthly toggle renders for me, the monthly figures of $9, $29 and $49 that independent reviews quote stay marked as unconfirmed in the table above, and the annual figures are the ones I would plan around. For how those compare across the field, the best Character AI alternatives ranking puts ChatFAI’s pricing next to a dozen competitors, and the best NSFW AI chat apps list covers what the adult switch costs elsewhere.