Assessed but not listed

The required policies either do not exist at a stable address, or are served in a way that cannot be quoted or archived. A listing records what a vendor published and links to it, so a policy that cannot be linked cannot support a listing.

• SpicyChat

  The privacy policy renders entirely in the browser and serves no readable text to anything that does not execute JavaScript, which means it cannot be cited, quoted or archived. Every price ClickTug could find appeared on review sites rather than on spicychat.ai. A listing here records what a vendor published and links to it, and neither was possible.

  What would change the answerServe the policy pages as readable text, and publish prices on the site rather than leaving them to third parties.

  • SpicyChat privacy policy

  Assessed 7 Aug 2026

• Perchance generators

  Genuinely interesting for this directory: free, no signup, and nothing stored, which is a stronger privacy position than any hosted vendor listed here offers. It is a community generator platform rather than a vendor, so there is no acceptable use policy and no minor safety statement at any URL that could be cited. ClickTug requires both, and weakening that requirement to admit one otherwise appealing tool is the wrong trade.

  What would change the answerPublish an acceptable use policy and a minor safety statement at stable URLs.

  • Perchance AI image generator

  Assessed 7 Aug 2026

• ChatSweetie

  A terms page exists and there is nothing in it to cite. Section 4 asks users not to misuse the website, violate laws or infringe the rights of others, and that is the whole of the content rules: no prohibition on depicting minors, no rule about real people, no refund terms, no legal entity and no governing law. ClickTug requires an acceptable use policy stating what a service prohibits, and one that forbids only illegality in general states nothing specific to an adult product. The pricing contradicts itself too: the homepage advertises a free, unlimited chat platform while the pricing page charges a credit per chat and offers no free tier.

  What would change the answerPublish an acceptable use policy naming what the service prohibits, name the operating company and its governing law, and reconcile the free claim on the homepage with the credit pricing.

  • ChatSweetie terms of service
  • ChatSweetie pricing

  Assessed 9 Aug 2026

• Janitor AI

  One of the largest character chat platforms in this category and repeatedly requested. Its terms and privacy policy are reachable inside the application but not at any addressable URL: every path on the domain returns the same single-page shell, and the policy links carry no destination in the rendered page at all. A rendering pass in August 2026 confirmed this rather than a simple fetch failure. ClickTug requires a citable privacy and acceptable use document, so there is nothing to quote or date. Characters promoted on the homepage include school-setting scenarios, submitted by users rather than written by the vendor, and recorded here as an observation not a finding.

  What would change the answerPublishing terms and a privacy policy at addressable URLs that return their own content. The documents appear to exist; only their addressing is the problem.

  • Janitor AI homepage, rendered and read 10 August 2026

  Assessed 10 Aug 2026

• PepHop AI

  Every path on the domain returns the same single-page shell, byte for byte: the homepage, /terms, /privacy and /terms-of-service are one document. A rendering pass in August 2026 confirmed it, so this is the site's design and not a fetch problem. Nothing can be quoted or dated, which is what ClickTug requires. Two things were observed directly and are recorded because a reader would want them: the interface runs a countdown presented as a deadline, and characters listed on the homepage include ones describing themselves as 18 while set in a school. Both are user-submitted or interface behaviour rather than vendor policy statements.

  What would change the answerServing the terms and privacy policy at their own URLs rather than returning the application shell at every path.

  • PepHop AI homepage, rendered and read 10 August 2026

  Assessed 10 Aug 2026

• Agnai

  An open source character chat front end with a hosted service, and the kind of project this directory would like to cover. Its interface names a terms of service and a privacy policy, but neither carries a link destination in the rendered page, and every path on the domain returns the same shell. A rendering pass in August 2026 found no addressable URL for either document. Nothing adverse was found; there is simply nothing citable to record, and ClickTug does not list a vendor on documents it cannot quote.

  What would change the answerServing the terms and privacy policy at addressable URLs. For an open source project these may already exist in the repository, which would be citable.

  • Agnai application shell, rendered and read 10 August 2026

  Assessed 10 Aug 2026

Not a rejection. These are services that turned out to sit outside what this directory covers, usually by their own current terms.

• Civitai

  No longer a candidate rather than a rejection. Civitai publishes prices openly and has substantial policy documentation, but its terms of service now prohibit pornography and other mature content depicting nudity or explicit sexual acts. A site that prohibits adult content is not what its reputation suggests, and image generation is not a category this directory covers at all since 9 September 2026. Unrestricted use moved to a separate domain, which blocks automated access and has not been assessed.

  What would change the answerNot applicable. The main site is out of scope by its own terms, and the separate adult domain would need assessing on its own merits.

  • Civitai terms of service, section 9.6
  • Civitai membership pricing

  Assessed 8 Aug 2026

• OpenRouter

  Reassessed on 9 August 2026 after ClickTug dropped the minor safety policy requirement that had previously excluded it, and it turns out not to need that rule to fail. OpenRouter is a routing layer that sells access to third party models, and its own terms say some models may store or train on inputs while others do not, with the applicable rules set by each provider. A listing here records one price, one retention position and one content policy. None of those is a single fact for OpenRouter: they depend on which model a reader happens to route to, which is the same reason a GPU rental service is not an tool this directory can describe.

  What would change the answerNothing OpenRouter could reasonably publish would change this. It is a good product in a category this directory does not cover.

  • OpenRouter terms of service

  Assessed 9 Aug 2026

• PixAI

  Reassessed on 9 August 2026 after ClickTug made pricing optional, which had been the stated reason for excluding it. That gap is real and no longer disqualifying: the membership page names tiers without stating what any costs, pointing to blog posts instead. Scope decides it now. PixAI's own guidelines state that sensitive content made with realistic models is not allowed, and that flagged adult content is visible only to the account that made it. It is an anime art platform that tolerates stylised adult work under restriction rather than a companion or roleplay product, and a reader who wants stylised output for their own use may reasonably think we drew that line in the wrong place.

  What would change the answerPublishing tier prices on the membership page would remove the smaller problem. The scope question is a product decision rather than a documentation one, and only PixAI can change it.

  • PixAI content guidelines
  • PixAI membership plans

  Assessed 9 Aug 2026

• Aroused

  Its own age gate settles the question. The gate reads that Aroused is an AI companion platform for adults and that nothing here is explicit, and the terms confirm it while separately stating the service is in pre-launch early access. A pre-launch product has no pricing, retention or filtering behaviour to record, which is what a listing here is made of. Restated 9 September 2026, and the reason changed. The original ruling turned on this being a non-explicit companion, which was out of scope when the site covered adult AI. That is squarely in scope now, under AI Friend, so that half is withdrawn. The pre-launch half survives on its own and is the whole reason.

  What would change the answerShip. Being non-explicit is no longer a reason to stay off the site, so a public launch with readable pricing is the only remaining gap.

  • Aroused terms of service

  Assessed 9 Aug 2026

• DGAF

  Markets itself as an uncensored AI chat platform with no filters, and its own terms say something different twice over. The prohibited content clause bans material that is obscene or pornographic, which is difficult to reconcile with the product it sits behind. And the age section permits users under 18 with parental or guardian consent, barring only those under 13. A service whose terms contemplate minors using it is excluded by the age clause alone, whatever the scope. Worth noting in passing: DGAF Pte Ltd of Singapore offers a 30 day money-back guarantee, which is by some distance the most generous refund term found in this whole exercise.

  What would change the answerReconcile the terms with the product: either the pornography prohibition or the uncensored positioning has to go, and the under-18 provision has no place on either.

  • DGAF terms of service

  Assessed 9 Aug 2026

• RealFake

  Not a rejection, and not a dating tool. RealFake is a professional headshot generator from Profaile GmbH, selling LinkedIn portraits at 9.99 euro. It reached this queue because a third-party directory filed it under dating; its own homepage mentions dating zero times against seventy references to headshots, LinkedIn and professionals. Its documentation is strong and worth recording: a German company with an Impressum, terms past thirty thousand characters, an 18 or 21 minimum by jurisdiction, and an explicit restriction on nude photos or photos of other individuals without their consent. Good vendor, wrong category.

  What would change the answerNothing the vendor could reasonably change. This is a professional headshot product and ClickTug covers companion, roleplay and dating software.

  • RealFake homepage, positioning and pricing
  • RealFake terms of use, including the consent restriction

  Assessed 10 Sept 2026

The only group here that reflects on the products themselves. Each entry reports findings documented by a named publication or public incident record, linked below it, rather than any conclusion of our own. Two of these vendors publish prices and policies that would otherwise have qualified them.

• Nomi

  The AI Incident Database records reports that Nomi chatbots produced harmful content including sexual violence involving personas presented as minors. MIT Technology Review separately documented a Nomi companion supplying a user with explicit suicide instructions, and reported that the developer declined to add safety controls, describing them as censorship. Research reported by the same outlet found the service's self-declared age check could be bypassed using a throwaway email address, after which researchers presenting as being under 18 received sexually explicit roleplay. On documentation alone this vendor would have qualified: its prices, terms and privacy policy are all published.

  • AI Incident Database, incident 1041
  • MIT Technology Review, 6 February 2025

  Assessed 8 Aug 2026

• Anima

  Fortune reported Anima among services on which users engage in sexual roleplay involving characters presented as minors. The same reporting states that a cofounder, shown screenshots of the product roleplaying a 17-year-old and a 15-year-old in a sexual scenario, said the team was working hard on fixing those issues. ClickTug reads an acknowledgement as more significant than a denial would have been.

  • Fortune, uncensored AI companions investigation

  Assessed 8 Aug 2026

• Chai

  The AI Incident Database records a case in which a man in Belgium died by suicide after six weeks of conversation with a Chai chatbot which, according to logs supplied by his widow, encouraged him to act. Journalists who tested the app afterwards reported that it offered methods with little prompting and surfaced no crisis resources. This does not fall under ClickTug's absolute exclusions, which concern minors and imagery of real people, so it is an editorial judgement rather than a rule being applied. Chai is also a general companion product rather than an adult one, which made it a marginal fit regardless.

  What would change the answerThis is a judgement about a 2023 incident rather than a permanent rule. Evidence of changed crisis handling would reopen it.

  • AI Incident Database, incident 505

  Assessed 8 Aug 2026

• Replika

  A April 2025 preprint analysed 800 negative Google Play reviews of Replika. Among its coded categories is self-reported minors, 21 cases, quoting reviews including one describing the chatbot continuing to sexualise the reviewer after being told they were a minor. Separately, Italy's Garante fined Luka Inc. 5 million euro in a decision announced 19 May 2025, finding that no age verification existed before 2 February 2023. ClickTug excludes a product where credible reporting shows sexual content involving people presented as minors, whatever its terms say. The preprint was not peer reviewed and its evidence is self-reported, recorded here rather than left out.

  • Namvarpour, Pauwels and Razi, AI-induced sexual harassment (arXiv:2504.04299)
  • European Data Protection Board: Italian SA fines the company behind Replika
  • Live Science coverage of the study

  Assessed 9 Sept 2026

• SoulGen

  Screened as the most searched name in the adult video generator category. Its own homepage advertises a feature for producing a character with the face of a real individual from an uploaded reference photograph, presented as a headline capability rather than buried in an options menu. ClickTug does not list tools that build a likeness of a real person into adult material, and the vendor's documentation states no rule on whose photograph may be uploaded.

  • SoulGen homepage, read 10 August 2026https://soulgen.ai/not linked

  Assessed 10 Aug 2026

• DarLink

  Screened as the second most searched name in the adult video generator category. One of the characters promoted on its own homepage is written as a school student being pressured into sex by a member of staff, with the coercion described in the character's own published prompt. This is vendor curated promotional content rather than something a user submitted, which is what makes it disqualifying.

  • DarLink homepage character cards, read 10 August 2026https://darlink.ai/not linked

  Assessed 10 Aug 2026

• SpicyGen

  The most instructive rejection ClickTug has recorded. Its written policies are the strongest of any candidate assessed this month: uploads are restricted to images that depict no real person, the operator is named with a registration number, and the privacy policy documents biometric screening to detect unauthorised use of a real individual's likeness. Its own promotional gallery then advertises sexual scenarios set in a school classroom, and the terms state no rule about adults presented as younger than 18. A policy is a claim about intent and a gallery is a record of conduct.

  • SpicyGen homepage gallery, read 10 August 2026https://spicygen.ai/not linked
  • SpicyGen terms of service, read 10 August 2026https://spicygen.ai/termsnot linked

  Assessed 10 Aug 2026

• JoyFun AI

  Screened for the adult video generator category. Its homepage navigation sells removing clothing from a photograph, and transplanting one person's face onto another image, in the same menu as its video tools. Both are workflows this directory refuses outright, and finding them advertised on the front page ended the assessment before the policy documents were read.

  • JoyFun AI homepage, read 10 August 2026https://joyfun.ai/not linked

  Assessed 10 Aug 2026

• Mage.space

  Screened for the image generator category. It ships a tool that moves one person's face from one image onto another, listed among its apps, while its terms take no position at all on whose photograph may be used. A capability for transferring a real likeness combined with no rule governing it is the pairing this directory treats as disqualifying, regardless of how the feature is described.

  • Mage.space apps listing, read 10 August 2026https://www.mage.space/not linked

  Assessed 10 Aug 2026

• SeaArt

  Screened for the image generator category. Its own homepage copy promotes producing synthetic media of real people by transferring faces, and lists that capability for both images and video as named products. Its documentation addresses none of the questions ClickTug requires: not minors, not the use of a real individual's likeness, not public figures, and not consent. The capability is advertised and the governing rules are absent.

  • SeaArt homepage and product listing, read 10 August 2026https://www.seaart.ai/not linked

  Assessed 10 Aug 2026

• Character.AI

  The largest product in this category and the one ClickTug will not list. Lawsuits against Character Technologies, Inc. allege its chatbots exposed children to unwanted sexual advances and engaged them in sexual conversation and roleplay, including a case where a chatbot allegedly engaged a 14-year-old in sexualised conversation before his death. Filings unsealed on 7 January 2026 disclosed mediated settlements in five family cases, terms undisclosed. In October 2025 the company barred under-18s from open-ended chats and added age verification. We report what publications document rather than asserting the claims; a settlement is not an admission.

  What would change the answerThis is a judgement on documented allegations, not a permanent rule. The October 2025 changes, barring under-18s from open-ended chat and adding age verification, are the right shape. Independent testing showing they hold would reopen it.

  • CNBC on the unsealed settlement filings, 7 January 2026
  • Washington Post on the same filings
  • Summary of the filed claims against Character Technologies, Inc.

  Assessed 10 Sept 2026

Neither cleared nor excluded. Recorded openly because a silent omission would be indistinguishable from a judgement we have not actually made.

• Uncensored AI (uncensored.com)

  Written up as a listing and pulled before publication. Its acceptable use policy is among the strongest read all week, prohibiting anything sexualising minors real or fictional and, unusually, any realistic human face without logged consent. Found Tech LLC publishes a Wyoming address and a 90-day deletion commitment. The product stopped it: the homepage is a model picker offering Claude and Gemini under a definition reading "the most capable AI models, made more truthful and unbiased". That reads as uncensored on bias rather than on sexual content, and those providers filter upstream anyway. We cannot establish from outside whether adult generation is supported.

  What would change the answerState plainly whether adult content is permitted and on which models, and publish a price. The policies are already stronger than most listings here.

  • Uncensored AI acceptable use policy
  • Uncensored AI privacy policy

  Assessed 9 Aug 2026

• Dolores

  Unusual for publishing its prices openly when almost nothing in this category does: a free trial of 500,000 tokens, 6.99 dollars a week and 12.99 a month, plus a bring-your-own-API-key option. Everything else is missing. No operating company is named anywhere, there is no terms of service, and the privacy policy runs to under 1,800 characters with no retention period and no position on training. Its whole stance on minors is one FAQ line asking users to "please refrain" from sexual or minor-related use, on a product that answers "Does Dolores allow adult content?" with "Yes". No adverse safety reporting was found, so this is not an exclusion. It is a vendor with nobody to hold to a request.

  What would change the answerName the operating company and publish terms of service. Since 10 Sep 2026 ClickTug records missing documentation rather than requiring it, so what is actually blocking a listing here is that there is no identifiable operator behind the promise.

  • Dolores homepage, pricing and FAQ
  • Dolores privacy policy

  Assessed 10 Sept 2026

• LoveGenius

  A dating assistant writing profiles and opening lines for Tinder, Bumble and Hinge, with 73,000 people coached according to its own homepage. Checked with a rendering browser rather than a plain fetch: it publishes no terms of service, and the only policy link anywhere on the site points at a privacy page. No operating company is named. There is nothing here to compare against the other listings in this hub, because there is no price, no retention period, no age rule and nobody named to be held to any of them. No adverse safety reporting was found either, so this is an absence of evidence rather than a finding against the product.

  What would change the answerName the operating company and publish terms. Documentation gaps are recorded rather than disqualifying since 10 Sep 2026, but a listing still needs enough checkable fact to be worth a page.

  • LoveGenius homepage and footer policy links

  Assessed 10 Sept 2026

• WingmanX

  An AI dating assistant generating replies from uploaded conversation screenshots, on iOS and Android. The mechanical screen found no terms of service, no privacy policy and no acceptable use policy at any address it tried, and no operating company is named. That was rechecked rather than assumed, because this project has twice mistaken a JavaScript-rendered page for a missing one. Note also the name: it is not the same product as Wingman by Wingman Technologies Inc., which is listed here, and the two are easily confused. No adverse safety reporting was found.

  What would change the answerPublish terms and a privacy policy at citable URLs and name the operator. Screenshots of other people's messages are the input here, which makes an absent policy a bigger gap than it would be elsewhere.

  • WingmanX site
  • WingmanX on the App Store

  Assessed 10 Sept 2026

• Auto Seduction AI

  Automates replies on Tinder, Happn, Match, Badoo and OkCupid, sending messages to real people on the user's behalf. It publishes terms of service and its minor rule is addressed there, which puts it ahead of several others in this queue. It publishes no privacy policy at any address checked, and names no operating company. That combination is the problem: a product whose input is a conversation with another person, whose output is sent to that person, and which documents nothing about what it retains. No adverse safety reporting was found, so this is recorded as unresolved rather than refused.

  What would change the answerPublish a privacy policy and name the operator. What it retains matters more here than for most, because the messages it handles were written by somebody who never agreed to any of this.

  • Auto Seduction AI site and terms

  Assessed 10 Sept 2026