AI × ART, DESIGN & MEDIA
Explore AI in Art, Design & Media
AI can supply material for a design, help search a recorded interview, or become part of an artwork. The creator decides what the work should communicate and which material belongs in it.
Examine a museum installation, a newsroom transcription case, and the source and authorship decisions behind a creative project.
Select a step to see what happens.
Creative brief
Define the audience and message
A brief gives the work a purpose. Source material, reuse terms, and the intended setting shape the available choices.
Visual alternatives
Compare possible compositions
Generated material can suggest alternatives. Judge hierarchy, readability, and relevance against the brief rather than selecting by novelty alone.
Selection and credit
Document the expressive decisions
Show what you selected, changed, and rejected. Record external sources and tool contributions alongside the finished work.
Illustrative design process. The composition is original explanatory artwork.
THE BASICS
Start with the audience and the brief
A school science-exhibition poster needs a clear message and readable information. A generated image might suggest a composition, but the designer must judge whether it communicates in the hallway where people will see it.
Generative models produce candidate images, sound, or text. A classification system can help organize a collection. A language model can assist research. A provenance standard serves a different purpose: recording information about a file's history and changes.[1][3]
Record what came from a source or tool, what you selected, and what you changed. Explain why you chose the composition and what revision taught you. Check generated claims and citations against their sources; NIST identifies confidently false output as a risk.[5]
APPLICATIONS
Where does AI fit in a creative process?
Follow the information through each task: the input, the model’s contribution, and the person responsible for checking the result.
Documented museum artwork
Building an artwork that responds to its setting
An artist wants a work to change over time and respond to its setting.
- The material
- In Refik Anadol's Unsupervised, MoMA describes collection data and site-specific environmental inputs.
- AI’s role
- The installation uses a machine-learning model within a generative artwork. The system produces changing imagery and sound.
- The human decision
- The artist's choices about source material, setting, scale, and presentation shape what the audience encounters.
The limitMoMA documents one artwork. Claims about creativity gains and permission to reuse source material require separate evidence.[1]
AP’s documented 2018 newsroom transcription rollout
Making recorded interviews searchable
Journalists need to locate relevant passages before selecting material for a story.
- The material
- Recorded interviews and events, with original audio or video retained for checking.
- AI’s role
- AP reported piloting and integrating Trint’s AI transcription service to produce searchable text linked to recordings.
- The human decision
- AP’s account says journalists review transcripts for accuracy before publication. Selecting and contextualizing material remains editorial work.
The limitA transcript needs checking against the recording. This dated deployment announcement is not an independent accuracy study or evidence about every current transcription tool.[6]
Published technical standard
Explaining where a media file came from
After a file is copied and edited, a viewer may struggle to understand its origin and changes.
- The material
- A digital asset and information recorded by compatible creation or editing tools.
- AI’s role
- C2PA Content Credentials can record signed statements about a file's origin, edits, and AI use. They help document the workflow around media, rather than generating the artwork.
- The human decision
- Read who signed the record and what it actually says. Combine this information with source checking and knowledge of the publisher.
The limitProvenance alone does not prove that a scene or claim is true. The standard also cautions against assuming that media without credentials is automatically untrustworthy.[3]
CASE STUDY
How Unsupervised connects data, system, and space
MoMA describes collection data and conditions in the museum lobby contributing to the installation.
MoMA exhibited Refik Anadol's Unsupervised from November 2022 to October 2023. Its account describes a model trained to interpret publicly available collection data, with changes in the museum lobby's light, movement, acoustics, and outside weather affecting the installation's imagery and sound.[1]
Separate three decisions when analyzing the work: the choice of collection, the transformation of that material, and the setting in which viewers encounter it. Each affects the account of how the artwork was made.
For a student project, use a brief you can revise and an audience you can describe. Document the material, the proposed transformation, and the presentation. Explain why you chose each element.
Follow the process Open a step to see what happens.
Choose the material
What are you interpreting, and what rights or context accompany it?
Define the transformation
What changes, and which decisions remain yours?
Consider the encounter
How should an audience experience and understand the result?
A process record should account for the source material, transformation, and viewing context, including your decisions at each stage.[1]
SKILLS & STUDY
What to learn first
- Develop a visual argument
- Learn composition, typography, contrast, and sequence. Explain what attracts attention first and how the work communicates.
- Practice critique and revision
- Compare alternatives against a brief. Record a criticism, revise a specific feature, and explain whether the change improved communication.
- Keep sources and credits
- Record the creator, source page, rights statement, tool use, and your own edits. Smithsonian Botany's policy distinguishes CC0-marked content from images with other conditions; check the actual item rather than assuming an entire website is unrestricted.
- Match the technical ambition
- A poster can begin with drawing and layout. Building an interactive system adds programming. Training a model adds data preparation, mathematics, computing resources, and evaluation.
PROJECT IDEAS
Projects that show your creative decisions
Choose a question, inspect the requirements, and keep a brief for later. Each proposal includes a way to check the work.
Save an idea in this browser or copy its brief to discuss with a mentor.
No account neededA design process with visible decisions
How could a poster communicate one school event to a specific audience?
- Intended output
- An original brief, selected alternatives, a refined composition, and an annotated process explanation.
Skills, sources and checks
- Starting knowledge
- Basic visual hierarchy and a drawing or layout tool. Coding is optional; the emphasis is on intentional design choices.
- Material to consider
- Your own work or individually checked public-domain assets, with source records and any tool terms kept alongside them.
- How to judge the result
- Compare readability, hierarchy, and relevance with the brief. Explain what changed after critique and credit every external ingredient.
A process-documentation project. Include the brief, revisions, critique, and credits alongside the final image.
A source and credit plan for a creative story
Could another person trace the claims and media ingredients in a short exhibition concept?
- Intended output
- A storyboard or layout proposal with a claim-to-source table, credit lines, and an explanation of planned AI assistance.
Skills, sources and checks
- Starting knowledge
- Research, note-taking, and distinguishing a historical fact from your interpretation of it.
- Material to consider
- Original collection records and assets with explicit reuse terms. A CC0-marked Smithsonian item is one possible starting point; inspect its record.
- How to judge the result
- Check each factual claim against a direct source and each ingredient against its rights statement. Explain any missing provenance instead of inventing it.
A planning project that keeps factual support and reuse permission in separate records. Attribution alone does not grant permission.
Document decisions, sources, and tool contributions. The legal discussion describes the U.S. Copyright Office's January 2025 report; permissions depend on material and context.
MENTOR SUPPORT
Plan a research workflow for a creative project
A research workflow could organize collection records, check factual statements, and prepare source notes for a creative project. Discuss the audience, sources, and review steps with a mentor in the STEMtiq Agentic AI Scholars Program.
Custom creative directions need a scope that fits the shared agent-building curriculum. The program does not replace specialist instruction in art or design.
5 weeks · 25 live hours · Grades 9-12
COMMON QUESTIONS
What students ask
Is generating an image the same as designing something?
An image is an ingredient. Design requires purpose, audience, visual decisions, and revision. Explain how you chose and changed material to communicate.
Can I use any image I find on a museum website?
Check the individual rights statement. Smithsonian Botany, for example, distinguishes CC0-marked content from other images with restrictions. Publicly viewable and reusable are different conditions. Keep the original record and relevant credit information.[4]
Does a Content Credential prove an image tells the truth?
No. It can provide verifiable information about origin and changes. You still need to examine the underlying claim and source. Missing credentials also do not automatically mean a file is false.[3]
Do prompts automatically make an AI image my copyrighted work?
The U.S. Copyright Office's January 2025 report does not treat prompts alone as sufficient in the systems it considered. It examines human expressive contributions case by case. That is a dated U.S. analysis, not a universal rule for every jurisdiction or future tool.[2]
What could mentor support add to a creative AI project?
It could help narrow a source-based workflow, make its tool permissions clear, and define checks for factual output. Ask about fit with the actual program. A mentor-scoped software workflow and specialist instruction in studio art are different commitments.
Discuss your creative research process.
Prepare a brief, an audience, and a few source records. A mentor can help assess which research or source-checking steps fit a software project.
Sources & further reading
- [1]Museum of Modern Art: Refik Anadol: Unsupervised (Exhibition: November 19, 2022–October 29, 2023)Return to guide ↑
- [2]
- [3]Coalition for Content Provenance and Authenticity: Content Credentials Explainer, specification 2.4 (Version 2.4; undated explainer)Return to guide ↑
- [4]Smithsonian National Museum of Natural History: Department of Botany: Image Use Policy (Undated institutional policy)Return to guide ↑
- [5]
- [6]
Sources inform this guide; their organizations do not endorse STEMtiq. Project ideas and diagrams are original editorial explanations.