Many creative teams adopted AI image tools the same way: someone learned to write good prompts, and the prompts became the process. That works for mood boards. It breaks down for production, and it carries a risk most leaders have not noticed.
A prompt is a weak way to specify layout, and a weak claim to authorship.
The Authorship Question Has an Official Answer
In January 2025 the U.S. Copyright Office published Copyright and Artificial Intelligence, Part 2: Copyrightability. Its conclusions include:
- "Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements."
- "Based on the functioning of current generally available technology, prompts do not alone provide sufficient control."
- Human authors are entitled to copyright in their expression that is perceptible in AI-generated outputs, and in creative selection, arrangement or modification.
The Office stresses that each case is decided individually, and this is not legal advice. The strategic signal is still clear: a process built purely on prompts produces assets with the weakest claim to protection.
Control Tools Change the Equation
Structural controls let people specify what text cannot. ControlNet conditions image generation on edges, depth, segmentation or human pose, so a sketch or a pose a person creates becomes an input the model follows. IP-Adapter adds an image reference with only 22 million extra parameters, letting a team show a look instead of describing it.
| Process | Speed to approved asset | Human expressive input |
|---|---|---|
| Prompt and re-roll | Unpredictable | Minimal |
| Sketch or pose, then prompt | Faster, steerable | The composition is human-made |
| Generate, then mask and edit | Targeted fixes | Creative modification is documented |
What This Means for Creative Operations
- Make composition a human step. Layouts, sketches and poses are where your team's expression enters the work.
- Keep the inputs. Archive sketches, masks and references alongside outputs. They are evidence of authorship.
- Train for direction, not just wording. The skill to build is controlling the model, not out-writing it.
- Review vendor tools for control features. A tool without structural controls pushes your team toward the weakest process.
The Business Case Beyond Copyright
Authorship is the headline reason to move from prompting to directing, but it is not the only one. Control changes the economics of creative work in three ways that show up in a production schedule.
- Fewer rounds. When composition is fixed by a sketch or a pose, review cycles focus on content and style rather than "move the product left", which a prompt can rarely guarantee.
- Consistency across a campaign. The same layout guide or reference image can drive a whole series, so assets look like they belong together.
- Targeted fixes. Masking and regenerating only the broken region preserves everything a client already approved, instead of re-rolling the whole image and starting review again.
What to Look for in Tools and Training
When teams evaluate AI image tools, prompt quality dominates the demos. It should be one criterion of several. Ask whether a tool supports structural guides such as edges, depth and pose, reference images, and region-level editing, and whether those inputs are saved with the project. A tool that only accepts text pushes a team toward the weakest process, whatever the quality of its model.
Training follows the same logic. The scarce skill is not writing longer prompts. It is the art director's skill of blocking a composition, choosing references and knowing which region to fix. Those skills transfer from traditional design teams, which is good news for organisations worried that AI makes their existing creative talent less relevant.
A Note on Evidence
If authorship may ever be disputed, the inputs are the evidence. Keep sketches, masks, references and the selection of final images with the delivered assets. The Copyright Office report treats human selection, arrangement and modification as protectable contributions. A project folder that shows those steps is a far stronger position than a prompt history alone.
What Changes for Creative Leads
Moving from prompting to directing changes how work is briefed and reviewed. Briefs can include a rough layout, not just a description, so the first round already has the right composition. Reviews can separate structure from style: approve the layout first, then iterate on look, the way a traditional shoot separates the storyboard from the grade. And the approval record shows exactly which human decisions shaped the final asset.
None of this requires new roles. It requires treating AI generation as a production step that takes direction, rather than a slot machine that takes wording. The principle is one creative teams have always known: the earlier a decision is made, the cheaper it is to keep.
Policy Language Worth Adopting
For organisations writing AI usage guidelines, a single clause captures most of this: "For assets intended for publication or client delivery, composition and key visual decisions must be made by a team member, using structural guides, references or edits, and the inputs retained with the project." It is short, it is enforceable in review, and it aligns everyday practice with both quality and the authorship position the Copyright Office describes.
Built In, Not Bolted On
In Vision Studio, our open-source generation app, masks, ControlNet guides and reference images live as layers on the canvas that the next render reads. Composition is placed by a person, not pleaded for in text, and the inputs persist with the project.
The Bottom Line
Prompting is a useful skill. It is not a production process, and it is not a strong basis for owning what you make. Teams that direct AI with human-made structure get faster approvals and a more defensible body of work.
Direct the model the way you would direct a photographer: decide the frame yourself, show what you mean, and fix the details that matter. The work gets faster and the result is more defensibly yours.
The hands-on guide to ControlNet, references and inpainting is on the Vision Studio blog at vision-studio-x.com/blog/stop-prompting-for-composition. Vision Studio is a free, open-source, local-first AI generation app built by Strategia-X. If you want a second pair of eyes on how your organisation adopts AI tooling, talk to us at strategia-x.com.
-Rocky
#AICopyright #IntellectualProperty #CreativeOperations #GenerativeAI #ControlNet #AIGovernance #StrategiaX #VisionStudio