Authorship as a Body of Decisions

Published On: June 28th, 2026Last Updated: September 15th, 20265383 words26.9 min readDaily Views: 1Total Views: 11

When people argue about authorship in AI-assisted work, they often begin with the finished output.

They look at the image and ask:

Who made this?

The question appears simple.

The image was rendered by a model.
A person entered instructions.
Perhaps the person selected one result from several.
Perhaps they edited it afterwards.
Perhaps the image belongs to a larger fictional world, campaign, visual identity, or body of research that existed before the generator was opened.

All of those facts may matter.

But the finished image alone rarely reveals enough to explain the human contribution behind it.

A polished output may have emerged from one vague sentence and a lucky generation.

Another image that looks equally effortless may have emerged from months of worldbuilding, established character design, visual research, repeated rejection, art direction, compositing, correction, layout, and editorial judgment.

The surfaces may appear similar.
The processes are not.

This is why I find it more useful to think of authorship not as a button press, a prompt, or a final file, but as a body of decisions.

Authorship becomes visible through the meaningful decisions a person makes, preserves, revises, and takes responsibility for across the construction of a work.

This does not automatically settle legal ownership.

  • It does not mean every decision carries equal creative weight.
  • It does not turn any amount of prompting into proof of authorship.
  • It offers a more useful way to examine what the human actually contributed.

That question sits at the centre of The Visual Register by Mithaq Praxis.

The output conceals the process

A finished image compresses its history.

It does not show:

  • the original intention;
  • the abandoned concepts;
  • the references considered;
  • the character rules applied;
  • the visual world already established;
  • the rejected generations;
  • the corrections;
  • the layout requirements;
  • the ethical boundaries;
  • the reasons one version was chosen over another.

The viewer sees the result.

They may not see the decisions that made the result belong to a particular person, project, organisation, or practice.

This is not unique to AI.

  • A reader sees a novel, not every deleted chapter.
  • A viewer sees a photograph, not every discarded frame.
  • A client sees the final logo, not all the research and rejected sketches.
  • A film audience sees the edited sequence, not every take.

Creative work has always involved invisible decisions.

  • Generative systems intensify the problem because the visible production event can happen very quickly.
  • The model renders in seconds.
  • The speed creates an impression that the entire act occurred in those seconds.
  • But the rendering event may be only one moment inside a much longer chain.

Or it may genuinely be nearly the whole chain.
The final image does not tell us which.

Not every decision is authorship

If authorship is a body of decisions, it is important not to inflate every interaction into a creative act.

Clicking:

  • square instead of portrait;
  • variation two instead of variation three;
  • generate again;
  • save;
  • upscale;

does not automatically establish meaningful authorship.

These may be necessary operational choices.
They may also be trivial.

The important question is not how many decisions occurred.
It is what kind of decisions they were.

Did the person determine:

  • what the work was meant to communicate?
  • which subjects belonged in it?
  • how those subjects related?
  • what visual world governed the image?
  • what should remain consistent?
  • what needed to change?
  • which references were relevant and why?
  • what made one result unsuitable?
  • what the final work required beyond generation?
  • when the work was complete?

A workflow containing fifty small adjustments may involve less authorship than one containing five substantial decisions.

Quantity is not the measure.
Meaning is.

The Visual Register should therefore not turn the creative process into a points system where more fields completed equals more ownership.
Its purpose is to preserve the decisions that shaped the work.

Intention is one decision, not the whole body

The human-written first line matters because it establishes where the work begins.

Four adult women try to balance together on one oversized pool float.

That scene seed contains:

  • the subjects;
  • the central action;
  • the shared event;
  • the playful premise.

It is a real human decision.

But intention alone does not complete the work.

The image may still require decisions about:

  • who is falling;
  • who is helping;
  • who is laughing;
  • who remains determined;
  • where the camera sits;
  • how the group remains readable;
  • whether the scene feels candid or posed;
  • what visual world surrounds them;
  • what boundaries apply;
  • which final version succeeds.

The first line protects the origin of the scene.
The wider body of decisions determines how that scene becomes a finished visual.

This distinction prevents two opposite errors.

The first error says:

The model rendered the image, therefore the human contributed nothing meaningful.

The second says:

The human wrote one sentence, therefore every visible aspect of the output belongs entirely to them.

Both are too simple.

The actual contribution has to be examined.

The Visual Bible contains prior authorship

A new image may inherit decisions made long before the present brief.

Suppose I write:

Farah and Zayd review a systems map in the Mithaq Praxis Research Library.

That sentence is short.

It does not contain every feature of Farah, Zayd, the library, the institution, its architecture, palette, material language, technology, emotional posture, or visual boundaries.

Those decisions already exist inside the Visual Bible.

They may include:

  • Farah and Zayd’s established appearances;
  • relative height;
  • attire;
  • jewellery;
  • recurring visual details;
  • the character of their interaction;
  • the Research Library’s architecture;
  • walnut, stone, paper, and muted metal;
  • restrained Islamic modernism;
  • credible near-future interfaces;
  • desert-facing institutional light;
  • the exclusion of generic cyberpunk styling;
  • the distinction between Mithaq Praxis and Algorithm Atelier.

When those decisions are carried into a new image, the brief may appear shorter.

The creative history behind it is not smaller.

The Bible represents accumulated authorship:

  • earlier conception;
  • visual differentiation;
  • repeated refinement;
  • corrections;
  • exclusions;
  • relationships among elements;
  • decisions about what the world means.

This is one reason prompt length is a poor measure of human contribution.

A short prompt backed by a deep visual system may contain more established direction than a long prompt generated automatically from generic adjectives.

Repetition does not strengthen authorship

If a character’s appearance has already been deliberately established, rewriting that description in every prompt does not increase the human contribution.

It increases clerical repetition.

The meaningful act was:

  • deciding the appearance;
  • refining it;
  • preserving it;
  • recognising when it drifted;
  • deciding whether it should change.

The same applies to:

  • a brand palette;
  • a recurring environment;
  • a logo rule;
  • a cultural boundary;
  • an attire system;
  • a visual motif;
  • an avoid list.

The Visual Register is designed to store these decisions so that authorship does not have to masquerade as repetitive data entry.

This is an important distinction.

Human-led does not mean forcing people to manually restate every established choice.
It means keeping authority over those choices with the person or team that made them.

The system can carry the decision.
It should not make the decision disappear.

Direction is a body of relationships

A scene is not authored merely by listing what appears inside it.

A person may request:

  • four women;
  • a pool;
  • an inflatable;
  • bright sun;
  • swimwear.

The result can still feel generic.

The more significant direction may lie in the relationships:

  • all four attempt to occupy one float rather than separate ones;
  • one loses balance;
  • one grips another’s arm;
  • one laughs too hard to help;
  • one tries to retain dignity;
  • their attention remains on one another;
  • the camera captures the shared event rather than presenting their bodies for display.

Those decisions create the scene’s meaning.

Similarly, a company may request:

  • four employees;
  • a prototype;
  • a research laboratory;
  • a screen.

The image becomes specific when the brief decides:

  • who demonstrates the prototype;
  • who records findings;
  • who identifies an error;
  • who compares the result against a diagram;
  • whether the environment feels credible;
  • whether technology supports the work;
  • whether the group appears collaborative or hierarchical.

Authorship often lives in the relationships among elements rather than in the elements themselves.

A generator can render the relationship.

The human can be responsible for conceiving and directing it.

References do not speak for themselves

Reference images are another area where authorship can become ambiguous.

A person may upload an image and say:

Make it like this.

What does like this mean?

  • the pose?
  • the lighting?
  • the medium?
  • the palette?
  • the architecture?
  • the clothing?
  • the camera position?
  • the emotional tone?

If the person has not decided, the model will choose which aspects to imitate.

The more deliberate act is to read the reference and state why it matters:

Use the overlapping group composition and casual body language. Do not copy the clothing or terrace furniture.

Or:

Use the restrained stone arches and natural daylight. Avoid the dense ornamentation and luxury-resort styling.

Or:

Reference the sepia line treatment and paper texture while preserving the established character canon.

The reference becomes part of authorship when it is interpreted and situated inside the project.
Collecting references can be meaningful research.

It can also be passive accumulation.
The difference lies in whether the person has decided what each reference contributes.

Rejection is a creative decision

A great deal of authorship exists in saying no.

A model may generate a striking image that the creator rejects because:

  • the emotional tone is wrong;
  • the relationship feels staged;
  • the architecture belongs to another world;
  • the product is inaccurate;
  • the cultural details are flattened;
  • the composition cannot support text;
  • the character no longer looks like herself;
  • the technology overwhelms the people;
  • the image is attractive but empty.

The discarded version may look more spectacular than the selected one.
Rejecting it can require stronger judgment than accepting the first attractive result.

This is why decision records matter.

A note such as:

Rejected because the large holographic display turned the Systems Laboratory into spectacle and displaced the people and work from the centre of the scene.

shows a meaningful decision.

It explains:

  • what the image did;
  • why that conflicted with the Visual Bible;
  • what the next version must preserve;
  • what must change.

The final image does not show that refusal.
The record can.

Selection is not automatically authorship either

Selection can be meaningful, but not every selection is equal.

A person may choose an image because:

  • it is the first one that looked attractive;
  • it resembles a familiar aesthetic;
  • it received the strongest immediate emotional reaction.

They may also choose because:

  • it best fulfils the brief;
  • the characters remain consistent;
  • the hierarchy supports the message;
  • the cultural details are coherent;
  • the image leaves appropriate layout space;
  • the emotional posture belongs to the project;
  • the result contains an unexpected but valuable visual idea.

The act appears the same:

Choose version three.

The reasoning differs.
Selection becomes part of the authorship body when the person can identify what they are preserving and why.

This does not require a long written defence for every casual image.
It becomes more important where the creator claims substantial authorship, professional responsibility, or organisational approval.

Revision shows the difference between taste and direction

A person may dislike an output and regenerate repeatedly.

That demonstrates preference.

A more deliberate revision identifies:

What should remain

  • the subjects;
  • the visual world;
  • the central action;
  • the colour relationship;
  • the environment;
  • the intended emotional register.

What should change

  • camera distance;
  • body language;
  • hierarchy;
  • lighting;
  • object placement;
  • architecture;
  • facial expression;
  • layout space.

This is design reasoning.

For example:

Preserve the night pool, warm terrace lanterns, four established women, and quiet companionship. Remove the duplicated pink-haired character, ensure all four women are distinct, reduce glamour posing, and give them natural conversational attention toward one another.

This is not simply asking for another image.
It is directing continuity and change.

The revision itself becomes part of the body of decisions.

Manual editing is visible authorship, but not the only kind

Some people treat manual editing as the point where an AI-generated image becomes human-authored.

Editing can indeed involve substantial creative work:

  • compositing;
  • painting;
  • redrawing;
  • masking;
  • reconstruction;
  • typography;
  • colour grading;
  • cropping;
  • retouching;
  • extending;
  • replacing generated elements.

These interventions can strongly shape the final work.

But manual editing should not become the only recognised form of human contribution.

A person may also contribute through:

  • worldbuilding;
  • visual-system design;
  • scene construction;
  • reference interpretation;
  • art direction;
  • selection;
  • rejection;
  • sequencing;
  • editorial placement;
  • written context.

Conversely, a minor manual change does not automatically transform the whole work into substantial human authorship.
Adding a border, adjusting saturation, or placing a signature over an output does not necessarily create a deep creative contribution.

Again, the question is not whether a human touched the file.
It is what the human decided and constructed.

Authorship can be distributed

Many visual works are not authored by one person.

A company campaign may involve:

  • a requester;
  • a brand lead;
  • a designer;
  • a photographer or illustrator;
  • a subject expert;
  • an editor;
  • a legal reviewer;
  • a model;
  • a retoucher.

A generative workflow may involve:

  • one person defining the Visual Bible;
  • another writing the present brief;
  • another producing and editing the image;
  • another approving the final use.

The body of decisions is distributed.

This makes the question:

Who made this?

less useful than:

Who contributed which decisions?

The Visual Register can help preserve that distinction.

For example:

Brand team

Defined the Company Visual Bible and locked product rules.

Requester

Defined the campaign purpose and audience.

Designer

Constructed the scene, composition, and visual hierarchy.

Generator

Rendered the initial image variations.

Designer

Selected, composited, corrected, and prepared the final layout.

Reviewer

Approved the asset for publication.

This account does not diminish the designer by acknowledging the model.
It does not exaggerate the model by erasing the human structure around it.
It describes the work more accurately.

Organisational ownership and creative contribution are not identical

In employment or commissioned work, the organisation may own or control the resulting asset under the relevant contract or policy.
That does not mean the organisation personally made every creative decision.

Similarly, an individual may make substantial creative decisions while working on material they do not legally own.
Ownership, authorship, attribution, and responsibility overlap, but they are not identical concepts.

The Visual Register should not attempt to settle these questions through interface labels.

It can preserve:

  • workspace ownership;
  • contributors;
  • brief history;
  • decision notes;
  • approved status;
  • source references;
  • declared execution method.

The legal meaning of those records may vary.
The practical value remains: the work’s construction is less opaque.

Copyright should not become a marketing promise

There is a strong temptation to market a process-recording tool as a way to establish copyright.

That would be irresponsible.

Different jurisdictions and institutions may treat AI-assisted material differently.

The significance of:

  • prompts;
  • selection;
  • editing;
  • arrangement;
  • transformation;
  • source materials;
  • human control;

may vary according to the facts and applicable law.

A product cannot honestly promise:

Build your Visual Bible and secure copyright over the image.

The Visual Register can do something narrower and more defensible:

Preserve the human decisions behind the work.

That record may be useful for:

  • provenance;
  • internal accountability;
  • client communication;
  • project history;
  • evidence of process;
  • future revision;
  • understanding contributors.

It may also help someone describe their creative role more clearly.
It is not a legal certificate.

This distinction should remain explicit.

A weak authorship claim may still accompany a meaningful experience

Not every generated image needs to become a claim of professional authorship.

A person may generate an image for:

  • play;
  • emotional exploration;
  • private enjoyment;
  • visual curiosity;
  • conversation;
  • a joke;
  • a personal fantasy;
  • experimentation.

They may have contributed very little beyond a broad request and selection.
The image may still matter to them.

Meaningful use does not require inflated claims.

The problem appears when someone:

  • contributes little;
  • conceals the extent of machine generation;
  • claims specialist labour they did not perform;
  • demands exclusive recognition over generic output;
  • treats the model’s decisions as evidence of their own mastery.

A healthier culture can allow people to enjoy generated material without forcing every interaction into a heroic narrative of authorship.
The Visual Register is intended for people who want to make and preserve more deliberate decisions.

It should not imply that casual generation is immoral.
It should resist the idea that casual generation and deeply directed construction are the same process.

Authorship is not a purity test

There is no useful purity in insisting that “real” authorship requires creating every component without assistance.

Creative work has always used:

  • tools;
  • collaborators;
  • editors;
  • references;
  • templates;
  • cameras;
  • software;
  • found materials;
  • libraries;
  • production systems.

AI is not identical to those earlier tools, but the existence of assistance does not automatically remove human authorship.

The more useful questions are:

  • What did the human determine?
  • What did the system generate?
  • What did the human accept, reject, or transform?
  • What continuity existed before generation?
  • What responsibility did the human take for the final result?
  • Is the process described honestly?

This avoids both romantic denial and technological exaggeration.

  • The model can make substantial visual contributions.
  • The human can make substantial conceptual, editorial, and structural contributions.
  • The final account should be capable of holding both.

The model also makes decisions

It would be misleading to say that the model merely follows instructions.
The model interprets.

It chooses, through its own generated process:

  • exact facial details;
  • folds of fabric;
  • reflections;
  • architectural transitions;
  • gestures not fully specified;
  • background objects;
  • colour relationships;
  • visual rhythms;
  • atmosphere;
  • composition within the remaining ambiguity.

Even a very detailed brief cannot determine every pixel.
The model contributes visual decisions.

  • Sometimes those decisions are generic.
  • Sometimes they are wrong.
  • Sometimes they are unexpectedly valuable.

Human-led work does not require pretending these contributions do not exist.
It requires the person to read them.

A model-generated detail may enter the work because the person decides:

I did not ask for this, but it strengthens the scene and belongs here.

That acceptance becomes a human decision about a machine proposal.

  • The origin of the proposal remains machine-generated.
  • The decision to preserve it remains human.
  • Both can be true.

Reading completes the authorship loop

The earlier article in this series argued that people must learn to read generated images.
This is directly connected to authorship.

A person cannot meaningfully claim to direct a work they did not examine.

Reading involves asking:

  • What did the model add?
  • What did it misunderstand?
  • What does the image now communicate?
  • Which details contradict the Visual Bible?
  • Does the relationship among the subjects match the intention?
  • What unintended cultural, emotional, or organisational claims appear?
  • Which accidental elements should remain?
  • Which must be corrected?

Human contribution is not limited to what appears before generation.
It also exists in what happens after the model responds.

The loop becomes:

intention → construction → generation → reading → decision → revision

Without reading, the process ends at output.
With reading, the person takes responsibility for what the output has become.

The brief is evidence of construction

A human-readable visual brief can preserve more authorship information than the final prompt.
The prompt may be condensed, platform-specific, or partly generated by the system.

The brief can show:

Purpose

Why the image exists.

Scene seed

What the human originally intended.

Visual Bible

Which established world, brand, characters, and constraints applied.

New direction

What was specific to this image.

Overrides

Where the scene deliberately departed from defaults.

Technical construction

How composition, hierarchy, light, format, and treatment were chosen.

Constraints

What must not change or appear.

This does not prove that every field represents an original creative contribution.
It makes the structure inspectable.

Someone can see how the work was conceived before it was translated into a generator’s language.

Decision notes should preserve reasons

A useful decision record should not merely say:

Version two rejected.

It should preserve enough reasoning to matter:

Version two rejected because the women appeared to pose independently for the camera rather than participate in one shared event.

Or:

The prototype remained accurate, but the laboratory drifted into generic cyberpunk styling and contradicted the Company Visual Bible.

Or:

The final crop removed the wheelchair user’s active engagement with the table, changing the representation from collaboration to passive presence.

These notes reveal:

  • what the image communicated;
  • what the creator or reviewer noticed;
  • why revision was necessary;
  • what standard governed the choice.

That is the kind of decision that contributes to a meaningful account of authorship and responsibility.

The record should not become proof theatre

Once a product can record decisions, users may feel pressure to manufacture an impressive history.
They may write elaborate notes to make a simple workflow appear more substantial.

  • An organisation may require unnecessary documentation to prove that staff performed enough “human work.”
  • A creator may believe that a long process log automatically strengthens a legal or artistic claim.

That would be a misuse of the system.

The Visual Register should not reward:

  • more prompts;
  • more rejected versions;
  • longer notes;
  • more fields;
  • more time spent.

A clean process may involve very few but highly consequential decisions.

The record should describe what happened.
It should not become a performance of difficulty.

The goal is legibility, not inflation.

Authorship may exist at the level of the series

A single generated image may contain limited human intervention.

The larger body of work may contain much more.

Consider a visual series built across months:

  • recurring characters are designed;
  • environments are established;
  • palette and motifs are refined;
  • narrative scenes are sequenced;
  • weak results are rejected;
  • visual treatments are coordinated;
  • images are selected and arranged;
  • captions and essays provide context;
  • the series develops a recognisable language.

Authorship may exist strongly in:

  • curation;
  • sequencing;
  • worldbuilding;
  • editorial structure;
  • thematic development;
  • visual continuity.

The individual image is only one component.

This is especially relevant for:

  • books;
  • visual essays;
  • campaigns;
  • fictional worlds;
  • exhibitions;
  • websites;
  • long-form blog series.

The Visual Bible helps preserve the structure that makes the images a body rather than a pile.

Arrangement can be a creative contribution

A person may not manually create every component of a page, campaign, or collection.
They may still make substantial decisions through arrangement.

They decide:

  • which image opens the series;
  • which one follows;
  • where the visual rhythm changes;
  • how text and image interact;
  • which motif recurs;
  • where silence or empty space is needed;
  • what the final sequence communicates.

This is not unique to generated work.

Editors, curators, designers, and directors have always created meaning through selection and arrangement.

The Visual Register should therefore preserve the relationship between:

  • individual briefs;
  • parent Visual Bibles;
  • campaigns;
  • series;
  • final placements.

This may belong to a later version of the software.

Conceptually, it matters now.
Authorship can exist in the construction of the whole.

A company’s body of decisions is also authorship-like

A corporate visual identity is rarely authored by one employee.

It develops through organisational decisions:

  • how the company represents its work;
  • which environments are shown;
  • what kinds of people appear;
  • how products are depicted;
  • what tone is acceptable;
  • what visual claims are considered truthful;
  • what is rejected;
  • how the brand changes over time.

The Company Visual Bible becomes a record of that collective construction.

Individual assets may be created by different people and tools.
They remain part of one authored visual system when the organisation has deliberately established and governed the relationships among them.

This does not mean a corporation should be treated as an artist in every philosophical sense.
It means visual identity is constructed through decisions rather than produced by a logo alone.

Responsibility follows decision-making

Authorship is often discussed as a claim to credit.
It also carries responsibility.

If a person claims meaningful control over a work, they should also be willing to answer for:

  • misleading representation;
  • cultural misuse;
  • brand inaccuracy;
  • exploitative references;
  • harmful stereotypes;
  • undisclosed fabrication;
  • privacy violations;
  • publication choices.

Human-led cannot mean:

I authored the image when it succeeds, but the model is responsible when it causes harm.

The person may not have intended every generated detail.
They are still responsible for reviewing what they choose to publish.

The body of decisions includes:

  • what was allowed to remain;
  • what was ignored;
  • what was considered acceptable;
  • what was released under the person’s or organisation’s name.

Authorship without responsibility becomes branding.

Provenance should describe contribution honestly

A provenance record should not reduce the work to:

AI-generated.

Nor should it disguise generation behind vague language such as:

digitally created.

A more useful description may distinguish:

  • human scene conception;
  • Visual Bible or brand system;
  • references;
  • model-assisted rendering;
  • selection;
  • editing;
  • compositing;
  • typography;
  • final review.

For example:

Scene, character continuity, and environment direction developed through The Visual Register. Initial image rendered with a generative model, then selected, corrected, composited, and prepared for publication by the designer.

Another project may honestly say:

Generated from a short exploratory prompt and published with minimal editing.

Both statements can be acceptable.

The goal is not to rank them morally.
It is to avoid presenting different levels of human contribution as though they were identical.

The product should preserve provenance without becoming a tribunal

The Visual Register can offer fields for:

  • execution method;
  • references;
  • tools used;
  • contributors;
  • revision notes;
  • human edits;
  • approval.

It should not automatically assign:

  • authorship scores;
  • originality percentages;
  • copyright likelihood;
  • creative-value ratings.

Those measures would pretend to solve questions the software cannot reliably answer.

The system should preserve facts and declared decisions.
People, organisations, and legal systems may interpret them differently.

A register records.
It should not pretend to be a court.

Human-led is a structural claim

Many products describe themselves as human-led because a person initiates the request.
That definition is too weak.

A workflow can begin with a person and still transfer nearly every meaningful decision to the machine.

For The Visual Register, human-led should mean that the product is structurally designed so the human retains authority over:

  • the original intention;
  • persistent continuity;
  • significant new choices;
  • overrides;
  • selection;
  • revision;
  • final approval.

The system can:

  • retrieve;
  • organise;
  • translate;
  • explain;
  • compare;
  • format;
  • adapt.

It should not silently decide the meaning and then present it as the user’s direction.

This is why the human-written scene seed matters.

  • It is why inheritance remains visible.
  • It is why the Visual Bible belongs to the user or governed workspace.
  • It is why system additions should be inspectable.
  • It is why revision reasons matter.

Human-led is not a marketing adjective.
It is an architectural requirement.

Authorship should remain possible without performance

The tool should not make users feel that they must narrate themselves as auteurs.

  • A designer may use The Visual Register simply because it helps maintain consistency.
  • A company may use it because staff briefs are chaotic.
  • A writer may use it to preserve fictional environments.
  • A casual creator may use it once and never create a Bible.

The software should support deliberate practice without demanding grand claims.

A person can say:

I directed this image carefully.

without claiming:

I manually created every visible element.

They can say:

This belongs to a world I built and maintain.

while acknowledging:

A model rendered much of the final visual material.

Accuracy is not diminishment.
It is the basis of credible authorship.

The body can be small

A meaningful body of decisions does not need to be enormous.

A person may make only a few important choices:

  • a clear scene;
  • an original relationship among subjects;
  • a carefully selected reference;
  • a deliberate composition;
  • a strong rejection and revision;
  • a final editorial placement.

Those choices may define the work.

Another person may produce a twenty-page prompt log containing little more than repeated aesthetic adjectives.
The length of the process does not determine its depth.

The Visual Register should therefore remain usable for small projects.
It should not equate seriousness with documentation volume.

A body of decisions is coherent, not necessarily large.

The body may precede the image by years

Some of the strongest visual decisions are made before the current tool exists.

  • A novelist may have developed a world over a decade.
  • An artist may have refined a character through hundreds of sketches.
  • A company may have accumulated a visual identity through years of design work.
  • A community may have cultural traditions that long precede the present project.

When a model renders an image from those systems, the generation does not become the origin of the visual identity.

The Visual Bible can help preserve this lineage.

It can distinguish:

  • inherited cultural material;
  • original project decisions;
  • commissioned design;
  • references;
  • generator contributions;
  • later edits.

This matters because AI conversations often compress everything into the moment of prompting.

The real construction may have begun long before.

The body continues after publication

Authorship does not always end when an image is posted.

A creator may later decide:

  • the image no longer fits the canon;
  • a cultural detail was wrong;
  • a character design needs revision;
  • the company’s visual rule should change;
  • the work should be removed;
  • the provenance note should be corrected;
  • the image belongs in a different sequence.

The Visual Bible evolves.

  • The decision history changes.
  • The creator takes responsibility for the work over time.

This continuing relationship can be part of authorship too.

The person is not merely the one who initiated generation.
They are the one who continues to govern what the image means within the larger practice.

The Visual Register should show the body without claiming to create it

The product does not manufacture authorship.
It cannot turn shallow participation into deep contribution merely by generating a beautiful record.

It can provide places to preserve:

  • the first intention;
  • persistent visual decisions;
  • scene-specific choices;
  • references and reasons;
  • overrides;
  • revision notes;
  • final approval;
  • execution method.

The substance must still come from the user or team.

A register can reveal a body of decisions.
It cannot invent one honestly where none exists.

This limit should remain part of the product’s integrity.

From prompt history to decision history

Most AI tools preserve a conversational history.

That history may contain:

  • prompts;
  • model replies;
  • generated images;
  • corrections.

It can be useful.

It can also be noisy.

A decision history is different.

It asks:

  • What did the human intend?
  • Which established rules applied?
  • What changed?
  • Why was a version rejected?
  • Which machine-generated addition was retained?
  • What was manually edited?
  • Who approved the final result?

This is more compact and more meaningful than preserving every exploratory exchange.

The Visual Register should not attempt to archive all thought.
It should preserve the decisions that explain the work.

The final image is one piece of evidence

When examining authorship, the final image matters.

So do:

  • the Visual Bible;
  • the brief;
  • the references;
  • the rejected outputs;
  • the revision notes;
  • the edit files;
  • the layout;
  • the series;
  • the written context;
  • the provenance statement.

No single artefact tells the whole story.
Together, they may show a coherent creative process.

This is the body.
Not a bureaucratic pile of proof.

A connected history of decisions through which the work became what it is.

The machine renders; the human remains accountable for the construction

The Visual Register is not being built to win an argument that every prompted image is human-authored.
That claim would be too broad and too convenient.

It is being built around a more demanding position:

Where people claim meaningful authorship, they should be able to identify the meaningful decisions they made.

Those decisions may include:

  • creating the world;
  • defining the visual identity;
  • writing the scene;
  • directing the relationships;
  • selecting references;
  • establishing boundaries;
  • rejecting outputs;
  • revising the composition;
  • preserving continuity;
  • editing the final asset;
  • arranging it within a larger work;
  • taking responsibility for publication.

The model may contribute substantially.
The human contribution should not be exaggerated.
It should not be erased either.

The purpose of the Register is to make the construction legible enough that we do not have to choose between those two dishonest extremes.

Authorship is not found inside the button

The button begins a rendering process.

It does not tell us:

  • who conceived the image;
  • what it belongs to;
  • which visual history preceded it;
  • what was rejected;
  • how the result was interpreted;
  • what was changed;
  • who accepted responsibility for it.

Those answers exist across the body of decisions.

  • That body may be deep or shallow.
  • Individual or collective.
  • Artistic, editorial, institutional, or commercial.
  • It may involve text, images, references, tools, models, and people.

The Visual Register gives it somewhere to remain visible.

  • Not as a guarantee.
  • Not as a score.
  • Not as a claim that the machine did nothing.
  • As a record that the work did not begin and end with Generate.

Because authorship is not a magic property produced by touching the button.
It is built through the decisions that give the work intention, continuity, form, and responsibility.

© 2026 • MITHAQ PRAXIS • CC BY-NC-ND 4.0 Unless Otherwise Stated.