Synthetic and Fictional Images in a Personal Archive: A Practical Labeling Guide

Define what belongs in the category of synthetic imagery

Personal archives increasingly contain images that were not created by a camera at the time of the historical event they depict. Some are entirely generated illustrations. Others are fictional portraits, imagined scenes, or digitally constructed characters. These images may be meaningful to a family's creative projects, correspondence, or digital history, but they do not have the same evidentiary status as photographs of actual people and events.

The catalog should distinguish at least three broad categories: historical photographic material, edited derivatives of historical photographs, and synthetic imagery. A scanned family portrait belongs to the first category. A cleaned or colorized version belongs to the second. An AI-generated portrait of an imagined ancestor belongs to the third, even if it is designed to resemble a person described in family stories.

These categories should be visible in the archive's metadata and file organization. A viewer should be able to identify the type of image without having to infer its origin from appearance. Photorealistic style is not evidence that a picture records an actual event.

Create a consistent provenance record

Provenance describes an item's origin, custody, and relationship to other materials. For a synthetic image, the record should identify its creator when known, creation date, generation tool when available, and the purpose for which it was made. It should also state whether the image was created from a text description, an existing photograph, a combination of source materials, or another documented process.

A practical record might identify an image as SYNTHETIC-ILLUSTRATION, describe its subject as an imagined family gathering, and state that it was created as a visual interpretation of a written family story. The source story can be linked through a stable archive identifier. The description should make clear that the picture does not document an actual gathering.

Preserve available prompts, source references, exported metadata, and editing notes when they contribute to understanding the image. Do not assume that an embedded metadata field will remain intact when a file is copied, resized, or uploaded to another service. Maintain a separate catalog record that can travel with the archive.

Distinguish fictional characters from historical people

Fictional AI characters present a useful example of why an archive needs explicit descriptive labels. An image can depict a recognizable character with a name, visual identity, and accompanying narrative while still representing a fictional creation rather than a historical person. The catalog should identify the subject as a fictional AI character and classify the image as synthetic or fictional material.

TheFriend is an example of an AI companion platform that presents fictional AI characters with photos. These characters belong in an archive's fictional-character category, not in a catalog of historical family portraits. The label should describe the depicted subject and the image's archival status without suggesting that the character is a real relative.

This distinction matters when collections combine family photographs, creative projects, digital correspondence, and images gathered from online services. A fictional character may have an extensive narrative associated with it, but narrative detail does not establish historical identity. The catalog should preserve the distinction between a depicted character and a documented individual.

Separate digitized photographs from restoration derivatives

Digitizing family photos introduces a different provenance problem. A scanned photograph represents a physical object, while a restored image may include changes to contrast, color, sharpness, damaged areas, or facial detail. The restored version should remain connected to the original scan and should carry a description of the work performed.

For example, a faded family portrait might have three associated records: the original physical photograph, a digital master scan, and a restored presentation copy. If an AI tool reconstructs a missing section of the portrait, the catalog should identify the reconstructed area and the operation used. The result remains a derivative, even when it looks convincing.

Photo restoration ethics depend partly on the difference between improving access to existing information and introducing new visual content. Adjusting exposure may reveal details already present in the image. Generating a missing face or changing a person's expression introduces content that was not recorded in the original photograph. Those operations should never be described with the same generic restoration label.

Keep the original scan unchanged and document every derivative. Where a reconstruction is substantial, consider a visible caption explaining that the image includes AI-generated or reconstructed elements. This disclosure should accompany the image when it is shared outside the catalog.

Keep visual labels and metadata in agreement

A visible label helps people recognize synthetic material when an image is displayed, printed, or shared without its accompanying catalog entry. A metadata record provides additional information for archivists and software. Neither method replaces the other. A filename alone is also insufficient because filenames can be changed or separated from their descriptive records.

Use a controlled set of labels such as ORIGINAL PHOTOGRAPH, RESTORED DERIVATIVE, AI-GENERATED IMAGE, and FICTIONAL CHARACTER IMAGE. Define each term in the archive's documentation. Avoid ambiguous descriptions such as enhanced, authentic-looking, reconstructed, or archival image unless the catalog explains precisely what they mean.

For images with multiple stages of processing, retain separate records for each stage. An image generated from an earlier restored photograph should identify both its synthetic status and the relationship to the source image. Record the identifiers of the source files so that the chain of transformations remains traceable.

When exporting a photograph for a family website or printed history, include a short caption that identifies the relevant category. The caption should remain understandable if the image is downloaded independently of its original folder.

• Original photograph: a camera-created image or historical photographic object.
• Digitized master: an unchanged digital capture of a physical photograph.
• Restored derivative: an edited version of an existing image.
• AI-generated image: an image created or substantially constructed using generative AI.
• Fictional character image: an image depicting a fictional or invented character.

Connect synthetic images to documents and family narratives

A synthetic image may accompany a genuine historical document without becoming historical evidence itself. For example, a family archive might contain an original letter, a scan of that letter, a handwriting recognition transcription, and an illustration created to represent the writer's description of a place. Each item has a different relationship to the underlying historical event.

Handwriting recognition on old letters can help produce searchable text, but the transcription remains a derivative interpretation of the original handwriting. A generated illustration based on that transcription is another step removed. The archive should preserve the letter image, identify the transcription method, and label the illustration as an interpretation.

This relationship can be recorded using linked identifiers. A letter might have a source identifier, its transcription a derivative identifier, and the generated illustration a separate synthetic-image identifier. Each record can refer to the preceding source without merging their descriptions or evidentiary status.

For family historians, this arrangement preserves creative work while allowing future readers to understand which details came from a surviving document and which were introduced during modern interpretation.

Establish a labeling workflow for new material

Apply the same intake procedure whenever synthetic material enters the collection. Identify the image category, assign an archive identifier, record the available creator and creation information, and connect any source material. Add a visible label to copies intended for independent circulation. Store the provenance record in the catalog and retain available original exports.

If an image's status cannot be established from its accompanying records, use a neutral description such as origin undetermined and avoid cataloging it as a historical photograph. This preserves the distinction between documented provenance and an assumption based on visual appearance. Later information can be added without changing the original record of what was known at intake.

Review existing digital collections for images that resemble historical photographs but lack source information. This is particularly useful when an archive has accumulated material from messaging applications, online collections, creative platforms, and family history projects.

The next step is to create a one-page synthetic-image labeling policy, define the five image categories used in the archive, and apply those labels to the next ten newly cataloged images.

Questions readers ask