Censorship Transparency Split Across AI Porn Generators

As of 2026-09-16, across the 23 platforms we track in AI Porn Image Generators, censorship transparency scores average 4.2 on a 10-point scale, ranging from 1 to 8. Of the 23 platforms scored on this axis, 13 rely on declared evidence pulled directly from a platform's own terms of service, only 3 have been independently observed in use, and 7 carry inferred scores because no usable policy documentation could be found at all. The axis measures how clearly a platform documents what it will and will not generate, and whether that documentation is detailed enough to be checked against actual behavior. The spread shows that most platforms describe their own rules on paper, but verification of those rules in practice remains rare across the category.

Few declared policies are backed by independent verification

The highest scores in the category belong to platforms with detailed, self-reported content policies. PornJoy AI and Sexy AI each score 8.0, the ceiling in this dataset, both on declared evidence. PornJoy AI's terms explicitly prohibit a documented list of illegal content categories, including non-consensual imagery, celebrities, incest and bestiality, and require documented consent for likeness use, backed by a lifetime ban for violations. Sexy AI's terms similarly ban CSAM, deepfakes, depictions of persons under 21, hate speech, bestiality and graphic violence. PornX scores 7.5, also declared, with terms that describe classifier or LLM-based moderation blocking prohibited requests, which the review notes aligns stated policy with an enforcement mechanism rather than leaving it as an unverified promise. By contrast, only 3 platforms in the entire dataset carry observed scores, meaning a reviewer verified behavior against stated policy with evidence. Sink In scores 2.5 on observed evidence: its terms require users be 18+ and list prohibited activities, yet the platform hosts NSFW-oriented checkpoints such as Babes and Realistic Vision without reconciling this against any visible explicit content policy. CreatePorn.com and AI-DreamGirls.com both carry observed scores of 2.0, in each case because no content policy or category restrictions could be found in the extracted terms despite explicit marketing claims.

Marketing claims of unrestricted content often outpace documented policy

Several platforms advertise permissive or "uncensored" generation while providing little accompanying documentation of limits. AliveAI.app scores 4.0: multiple models, including Creative v2, Realism, Krea2 and Anima 1.0, are marketed as "Uncensored," yet the terms reference an Acceptable Use Policy, Blocked Content Policy and Underage Usage Policy that are not detailed on the page itself, creating a gap between the marketing language and what a user can actually verify. AI-DreamGirls.com, scoring 2.0, markets "no censoring, create and clone nude images," but its terms of service contain no explicit content policy or list of prohibited material. Lovescape, scoring 1.5 on inferred evidence, advertises "Unfiltered AI Porn Images & Videos" on its homepage, but the terms page content was not accessible beyond its title, leaving no documented limits to check against that claim. Not every permissive platform follows this pattern: Unstability.ai scores 7.5 on declared evidence precisely because its terms explicitly allow 18+ users to disable a safety filter, which the review notes matches the site's own stated "minimal content restrictions" positioning rather than contradicting it. The distinction across these cases is not whether a platform allows explicit content, but whether its marketing claims are matched by terms a user can actually read and verify.

Inferred scores mark platforms with no discoverable policy

Seven platforms in the dataset carry inferred scores, the category's lowest-confidence method, applied when reviewers could not locate usable policy documentation at all. AI-PORN.AI and NovelAI each score 2.5: no terms of service were found detailing content restrictions for AI-PORN.AI beyond an age check and takedown process, while neither NovelAI's terms nor its privacy policy excerpts define moderation limits or prohibited categories. CreateAIBBW.com and CreateAIMilf.com each score 2.0; the former's terms page returned only navigation links with no visible policy, and the latter's terms page content was empty in the extracted data despite the site carrying 2257 compliance links. PixAI, also scoring 2.0, had no documented content policy, age verification, or consent policy despite operating as an AI porn generator. Nolo scores the lowest in the entire dataset at 1.0, with no terms of service or content policy found at all, leaving any claims about its generation limits completely unverifiable. These inferred scores do not mean the platforms lack internal rules; they mean no external documentation exists for a reviewer, or a user, to check those rules against.

What to check before paying

Given that only 3 of 23 platforms in this dataset have observed rather than declared or inferred transparency scores, users should not assume a platform's marketing language reflects its actual moderation practice. Before subscribing or paying for credits, it is worth checking:

• Whether the terms of service list specific prohibited content categories, rather than a general statement about "illegal content"
• Whether any moderation or filtering mechanism (such as classifier-based blocking) is described, not just claimed
• Whether age-verification or consent requirements for real-person likeness are documented, particularly on platforms offering realistic generation
• Whether marketing terms like "uncensored" or "unfiltered" are reconciled anywhere in the terms, or left unaddressed
• Whether the platform's terms page is even accessible and populated, since several platforms in this dataset returned empty or navigation-only terms pages

Because 7 of the 23 scores here rest on inferred evidence and 13 rest on the platform's own declared documentation, this dataset reflects what is written and discoverable, not a guarantee of what any platform enforces in practice.