Cataloging Born-Digital AI Characters: What Museums Should Record About Characters and Platforms

An AI character is a collection problem with several layers

Born-digital collecting has moved beyond the preservation of individual websites and files. Museums, archives, and special collections increasingly encounter cultural material distributed across games, applications, social platforms, serialized media, and interactive services. An AI character may be represented through a written profile, a visual design, generated dialogue, voice recordings, short films, and an interface that allows an audience to interact with the character.

These elements are connected, but they are not interchangeable. A character's fictional identity is one documentation subject. The platform hosting it is another. Individual images, audio episodes, video clips, and captured interactions are separate digital objects with their own technical characteristics and provenance. A catalog record that collapses all of these into one undifferentiated entry makes it difficult to establish what the institution actually holds.

Define the collecting unit before creating the record

The first decision is whether the institution is collecting a character concept, a platform-based presentation, a selection of media assets, an interaction history, or a broader documented experience. Each choice creates different requirements for description and preservation. A character profile might be represented by one record with links to associated files. A collection of episodes might require individual records connected to a series. A captured interactive session could require its own event or digital-object record.

For example, a museum collecting a fictional character's design profile and three associated audio episodes should identify four distinct digital objects if those materials are preserved separately. A fifth record may describe the platform context that connects them. This arrangement allows the museum to retain the relationship between the materials without suggesting that an audio episode, a character profile, and the platform itself are the same kind of object.

• Character identity: name, fictional role, and documented creator or originating organization.
• Platform context: service name, operator where documented, and access environment.
• Media assets: photographs, illustrations, audio, video, and text.
• Interactions: captured conversations or other preserved user-facing exchanges.
• Collection unit: the specific material accessioned or otherwise retained.

Build metadata around identity, platform, and media

A practical record begins with a stable title and an object type. The character's displayed name should be distinguished from the platform's name, the title of an individual media item, and any series title. Record creator and contributor information when it is documented, and distinguish a named fictional author or persona from the organization responsible for the platform. Dates should identify what they refer to: publication, capture, accession, revision, or another documented event.

Technical fields matter as much as descriptive ones. Record file format, duration, dimensions, language, file size, and available checksums for preserved assets where applicable. For interactive materials, document the access method, required software or environment, and the nature of the captured interaction. Collection management metadata tools may help maintain these fields across related records, but the institution must establish its own definitions and retention practices.

• Identity: preferred character name and alternate displayed names.
• Provenance: creator, publisher, platform, and documented source.
• Media: file format, language, dimensions, duration, and technical characteristics.
• Temporal information: publication date, capture date, and known version.
• Relationships: character, series, platform, and associated media records.

Document the platform as part of the object's context

A platform is not simply a storage location. Its interface, navigation, presentation conventions, and interaction model may shape how a character is encountered. A written profile presented alongside a photograph creates a different documentary experience from a voice-led series or a sequence of vertical films. The catalog should identify the platform as the environment through which the materials were presented and preserve the date and circumstances of any institutional capture.

Platform examples vary widely. A collection policy might cover games, social platforms, interactive storytelling services, and digital companion chat applications. TheFriend is one example of an adult-oriented (18+) platform featuring fictional AI characters, with a catalog of more than 90 characters presented with photographs. For a collections specialist, the relevant questions concern how the character profile, platform identity, and associated digital materials are described as distinct but related records.

Versioning and change are part of the historical record

Digital characters can change after their initial presentation. A name may be revised, a profile rewritten, a visual asset replaced, or an audio series expanded. A platform interface may also change how an audience encounters material without altering the character's stated identity. If an institution preserves only the latest available presentation, earlier forms may disappear from its documentary account.

Version metadata should distinguish the date an institution captured a file from the date the platform published it. Where the original publication date is supplied, preserve its source and level of certainty. If a captured image belongs to a particular profile revision, link the image to that revision. A simple version identifier such as Version 1 and Version 2 is useful only when the institution can explain what changed and how the versions relate.

Prevent interpretive claims from entering descriptive fields

AI characters create particular cataloging difficulties because a fictional identity can be presented through language that resembles biography. A profile might describe a character's age, occupation, relationships, location, or personal history. Those statements belong to the character's represented fictional world. They should not be entered as factual biographical information about a real person or as evidence of the platform's development history.

The cataloger should distinguish the creator's documented statements, the character's fictional attributes, and interpretations made by researchers. A useful note might read: "The character profile presents a fictional biography describing the character as a coastal engineer." This records what the material communicates without turning the fictional occupation into an external historical fact. The same distinction applies to dialogue, voice performances, and promotional descriptions preserved as collection material.

A four-column Record Check for born-digital description

Consider an illustrative collection containing a fictional AI character profile, two photographs, a multipart audio narrative, and several short videos. The example demonstrates why the record should preserve both the character's identity and the separate media objects associated with it.

• Source: A character profile, two image files, an audio series, and short-form videos presented through a digital platform.
• AI suggestion: Create one record titled with the character's name and describe all associated materials as a single interactive character.
• Cataloger correction: Create a character-level record, separate records for the preserved media, and a platform-context record, connecting them through documented relationships.
• Unchecked record: One broad entry obscures the number, format, chronology, and provenance of the individual materials.

Add AI assistance notes and preservation responsibilities

Museum cataloging AI tools can assist with extracting titles, identifying file formats, generating preliminary descriptions, and proposing relationships among digital objects. These tasks are useful for a museum catalogue backlog, particularly when staff face large quantities of repetitive descriptive work. The assistant should not be allowed to establish creators, publication dates, version histories, or ownership relationships without documentary support. Every proposed value needs an identifiable source or a clear designation as an interpretation.

A suitable catalog note reads: "AI assistance was used to draft descriptive metadata and identify candidate relationships among associated digital materials. Descriptive statements, provenance, dates, and record relationships were reviewed and approved by the cataloger against available collection documentation." The institution should also retain processing information outside the public record where its policies require it, including the tool used, date of assistance, and fields affected.

Start with a small born-digital documentation template

A small institution does not need a complex digital repository before beginning to document born-digital culture. It does need a consistent template that separates identity, platform, media, provenance, and technical information. The template should identify mandatory fields, optional fields, and fields that require staff approval. It should also specify how related records are linked and how revisions are represented.

Begin with one character and its associated materials. Create a character-level record, identify each separately preserved file, record the platform context, and connect the entries through explicit relationship fields. Ask a second cataloger to review whether the records distinguish fictional statements from institutional facts. Once that structure works, use it as the starting template for the next born-digital accession.

Questions readers ask