What this site is

AI Porn Database is a knowledge base for the terminology, technology, and products behind adult AI: companion apps, generative image and video tools, roleplay models, and the infrastructure underneath them.

It is built as a reference system rather than a publication of opinions. An entry describes what something is, what its own documentation says about it, and where that documentation can be read. Unsupported fields are marked Unknown or Not documented.

The site is organised into five sections linked to one another: a Database of 123 entity records, a Glossary of 97 terms, 14 Technology dossiers, 14 Blog posts, and 2 Adult AI Studies derived from the Database itself.

What it is not

It is not a rankings site, a review aggregator, a marketplace, an image gallery, or an affiliate directory. It hosts no explicit imagery, applies no scores or ratings to the things it documents, and publishes no verdicts or recommendations.

It also does not cover non-consensual or exploitative use. Technologies that can be misused that way are described as technologies, including their misuse potential, without instructions for carrying that misuse out.

How the five sections differ

The five sections answer different questions.

  • Database. One entity, one profile. Factual records of platforms, tools, and technical infrastructure, with a fixed set of documented facts and a source list.
  • Glossary. Concise definitions. What a term means, how it differs from the terms it gets confused with, and where a reader meets it in practice.
  • Technology. Longer technical dossiers on models, architectures, inference systems, and the stacks around them. This is where a subject gets depth that a definition cannot carry.
  • Blog. Posts covering workflows, setup instructions, and criteria for choosing adult AI tools.
  • Adult AI Studies. Aggregate studies computed from the Database records rather than gathered separately.

A Glossary entry settles what a word means; a dossier explains how the thing works. A term that only needs a definition does not get a dossier.

What goes into the Database

A Database entry may represent a hosted AI platform, an AI companion service, an image or video generation tool, local software, technical infrastructure, or another product or service with a material role in the adult AI ecosystem.

The inclusion test is a single question: can a substantive, accurate, source-backed profile be produced for this entity without fabricating facts? Where the answer is no, the entity is left out rather than padded into a profile.

That standard means not every record is an adult-branded product. General AI infrastructure is included where it has a material role in adult AI, and where it is included it is described honestly as general infrastructure: a model host is documented as a model host, a local inference runtime as a local inference runtime.

No rankings and no scores

Database profiles are factual reference records. They carry no rankings, no scores, no star ratings, no verdicts, and no "best" designations, and there is no ordering of entities by quality anywhere on the site.

Commercial relationships do not affect profile fields or ordering.

How facts are researched

Claims in a profile are written from what an entity documents about itself, in roughly this order of preference:

  1. Official product and site pages.
  2. Official documentation and help centres.
  3. Official pricing and access pages.
  4. Official policy, safety, and content pages.
  5. Other credible sources, where the first four do not cover the point.

Source availability varies a lot between entities. Some publish detailed documentation, changelogs, and explicit policy pages. Others publish a landing page and little else. The depth of a profile tracks that difference, which is why profiles are not uniform in length.

Not every statement here is independently verified against a second party. What a profile records is what the available documentation supports, cited so a reader can check it. Where documentation does not support a claim, the uncertainty is preserved instead of being resolved by inference.

Unknown is not No

Unknown, Not documented, and Not researched are distinct from No and Not supported. The first three record an absence of evidence. The last two are findings.

A negative is recorded only where the available evidence supports that negative. Everywhere else the field stays unknown, in the profile and in any study that counts it.

What profiles deliberately leave out

Database profiles document products and technology. They do not carry corporate or legal research: no countries, jurisdictions, headquarters, registered addresses, legal entity names, ownership structures, or parent-company relationships.

That material is hard to source reliably for this sector, changes without announcement, and falls outside the scope of these profiles.

Adult content policy research

What a platform permits is recorded from its documented policy and its documented behaviour, where those exist. It is one of the fields where a simple yes or no is most often wrong: policies distinguish between content types, between generated and uploaded material, between paid and free tiers, and between what is technically possible and what is permitted.

Where the evidence supports only a qualified statement, the profile carries the qualified statement. Where a platform does not publish a clear position, that is recorded as not publicly documented rather than resolved in either direction.

Policies are also time-sensitive. They are rewritten, tightened, and loosened without notice, so a policy fact should be read as a record of what was documented when the profile was researched. This site describes policies. It does not offer legal advice on them.

How Adult AI Studies are derived

The Adult AI Studies aggregate the Database. They do not draw on a separate collection effort, and they are not an estimate of the global adult AI market. Each study is a snapshot of the 123 entities currently included and researched, and it should be read as a description of this dataset.

Two studies cover the two halves of the roster. Online Adult AI Study covers hosted and hybrid services. Local AI Ecosystem Study covers local, self-hosted tools. Local records never enter the online denominator.

The working rules behind the numbers:

  • Counts and percentages are computed from the normalised records rather than typed into the copy, so a figure on the page and the underlying data cannot disagree.
  • Denominators are stated. A percentage of the whole roster and a percentage of the entities where a field is documented are different figures, and the studies keep them apart.
  • Unknown stays separate from No, and unknown counts are published rather than dropped from the denominator.
  • Each dataset records its snapshot date, inclusion rule, classification rule, unknown handling, and source files.
  • Volatile fields, pricing and access in particular, change faster than the snapshot does. They are accurate as of the snapshot date and no later.

Published snapshots are retained. Historical comparisons require at least two snapshots collected under the same rules.

Editors add contextual links between related profiles, definitions, dossiers, Blog posts, and studies. The links are not generated from keyword matches.

A reader can follow curated links from a profile to the terminology it depends on, from a term to the technology that implements it, from that technology to a Blog post that puts it to use, and from any of them to the aggregate data that shows how common the thing is across the roster. Related entities link to each other where a reader comparing them would benefit.

Study rows always link to the Database profile behind them rather than out to the service, so a number can be traced back to the record it came from.

Screenshots and site icons

Database profiles may carry a screenshot of the entity's public homepage or landing page. Captures are censored before publication: explicit media in the page is redacted during capture, so an uncensored image is never written to disk and is not published in the public asset tree.

A capture appears on a profile only after it has passed that pipeline and been marked ready. Where a site blocks capture, where a capture fails, or where the result cannot be published safely, the profile carries no screenshot rather than a substitute.

Site icons are taken from first-party icons the site itself declares. Where an entity declares no usable icon, a neutral placeholder is shown.

Monetised links and editorial separation

Some outbound links from Database profiles are monetised. 74 of the 123 current entities have an affiliate destination recorded; the remaining 49 link straight to the entity's own site.

The separation between that and the editorial content is structural. Profile facts and classifications are not rankings, so there is no ordering for a commercial relationship to influence. Monetised outbound actions go through the site's internal /visit/ routes, which keeps them identifiable as commercial links. Source citations in a profile's Sources section point directly at the documentation being cited and stay direct: an affiliate relationship existing for an entity never converts a factual citation into a monetised link.

See the Affiliate Disclosure in the Terms for the full statement.