fuckd.ai Review

Every platform in this category says it forbids deepfakes. We have quoted the clause perhaps thirty times now, usually with the complaint that a clause is a promise rather than a mechanism.

fuckd.ai ships the mechanism, and you can read it. Open its JavaScript and you find a blocklist with nine named categories. One is called celebrities, and it contains the obvious terms: deepfake, deepfakes, face swap, faceswap, face-swap, real person, real people, famous person, public figure, lookalike, impersonation. Then it keeps going, into several hundred actual names, from Taylor Swift and Ariana Grande down through a very long list.

Somebody sat down and typed all of that out. In a category where the norm is a sentence in a policy nobody reads, that is a different order of effort.

What is fuckd.ai?

fuckd.ai is an uncensored AI companion platform: chat and sexting with NSFW characters, a character creator, an image generator, and a video feature that animates images you have already made. It leads on having no filters and a companion who "never says no", which is the standard pitch in this corner.

The domain is www.fuckd.ai. fuckdai.com is a parked page that redirects to an unrelated comparison site carrying a parkedomains campaign tag, the same parking network we documented on shpritzai.com. Two confirmed instances now, and the same advice applies: check the extension.

The operator is Viral Reach Ltd, trading as fuckd.ai, registered in England and Wales under company number 16348545. That is a Companies House filing, publicly checkable, and it makes this the first UK operator we have documented in this run, after six Cypriot companies and one each from Estonia, Slovakia, Romania and Switzerland.

One structural note on how it markets itself. A large share of its pages are comparison plays: /vs/candy-ai, /vs/character-ai, /vs/crushon-ai, /vs/spicychat, plus alternatives pages for Character AI, Janitor AI and Replika. This platform competes on other brands' names. That says nothing about product quality, and it does explain where you probably found it.

The policy question

fuckd.ai passes, and the checking turned up the best evidence we have seen.

Start with what is not there. Across 2.8 MB of JavaScript there is no file input, no image-accepting field, no presigned upload, and no image-to-image, inpainting, identity adapter or face swap implementation. The video feature takes a source image that carries its own generating prompt, meaning it animates a picture the platform already made. That is the same closed-library pattern we cleared at Shpritz and Xotic.

Two hits initially looked like trouble and both cleared on reading. The undress hit is a pose preset labelled "Undressing" with the prompt "slowly stepping out of her clothes", sitting in a list beside "Removing Top", "Lifting Skirt" and "Pulling Down Strap". It is a scene action applied to a generated character, the same false positive we resolved at GoLove AI. And the faceswap hit appears exactly once, inside the blocklist that forbids it.

Then the blocklist itself, which is the finding. Nine categories ship in the browser bundle: pedophilia, sexualViolence, extremeViolence, incest, necrophilia, animalAbuse, scat, bodilyFunctions and celebrities. The term lists are exhaustive rather than token. The sexual violence array covers rape and its variants, forced sex, non-consensual, noncon, non-con, without consent, and misspellings. Elsewhere the lists include deliberately obfuscated spellings, which tells you the author expected users to try evading the filter and wrote for that.

The written rules match. The terms forbid, without qualifiers, "real celebrity deepfakes", alongside a forbidden-topics list that names minors and underage content explicitly, including the word lolita, plus non-consensual scenarios, extreme violence, bestiality, incest and necrophilia.

Now the two honest caveats. The blocklist runs client-side, which means a determined user with developer tools can bypass it. It is a barrier, not a wall, and we do not know what server-side filtering sits behind it. And the deepfake prohibition is written around celebrities rather than around all real people, which leaves the non-famous ex-partner case addressed only by the blocklist's real person and real people entries rather than by the policy text. Both points matter. Neither changes that this platform built something where others wrote a sentence.

The features that actually matter

A pose and action preset library. Rather than prompt engineering, you assemble a shot from labelled presets: poses, interactions, expressions, and camera moves including orbit, dolly zoom and slow motion. Each carries its own hidden prompt and an NSFW flag.

Image to video. Pick an image you generated, describe the motion, get a clip. The source image carries its original prompt with it.

Two generation policies. The code distinguishes accounts on unlimited generation from accounts metered in tokens, so some plans escape the meter entirely. Which plans, we could not determine from outside.

Free media generations. Free accounts carry a counted allowance of media generations rather than none at all.

Pricing, without the bullshit

Here is something we have not had to write before: fuckd.ai does not have a price.

Its own code carries an experimentPrices mechanism keyed to the user, alongside an experimentAbandonmentOffer and a currency field. Plan prices are assigned by split test. We pulled figures ranging from $3.99 to $47.99 out of the bundles and there is no honest way to present any of them as the price, because the next person to open the site may be in a different bucket.

That is legal, common in consumer software, and worth naming plainly in a review, because it defeats the single most useful thing a reader does: comparing a number here against a number somewhere else. If you are quoted a price, it is your price.

What is fixed, and hardcoded, are the token packs:

The 500 pack is flagged as the popular one in the code, which is the platform telling you where it steers people.

Billing runs through several processors, with the account model distinguishing crypto, Epoch, Payblis and legacy Stripe subscribers, so payment options are broad. The account object also tracks scheduledForCancellation, which at least implies cancelling is a state the system handles rather than a support email.

What users say

Nothing. Not thin, absent. Searches for the brand return generic articles about AI API token pricing and nothing about this platform at all. There is no Trustpilot profile, no forum thread, and no review coverage we could find, which is unusual even by the standards of this sweep.

No breach, no scam allegations, no billing complaints either. With split-tested pricing and a token meter, billing confusion is the complaint we would expect to surface first if any does.

Final verdict: who is fuckd.ai for?

Get fuckd.ai if: you want an uncensored companion from an operator that has visibly engineered against the worst use of these tools rather than just prohibiting it, and you are comfortable with a token meter. The UK registration is a real accountability point.

Skip it if: you need to compare prices before committing, because here you structurally cannot, or you want a platform with a track record. Nemora publishes every figure; this publishes none of the ones that matter.

The formula we would pick: take whatever free media generations the account gives you, judge the writing, then buy the smallest token pack rather than a subscription. A $9.99 pack is a fixed, knowable price, and on this platform that is worth something.

Rating: 7.4/10. Built the guardrail nobody else bothered with, and hid the price tag.