← Encyclopedia
Skill
Introduction
A generative system may produce an apparently finished advertisement containing incorrect writing, uneditable details, or material whose origin the team cannot explain. A designer's responsibility therefore continues after output appears rather than ending with selection of the most attractive version. This article follows a fictional educational-tools advertisement from permissible inputs through review and delivery, focusing on provenance, maintainability, and truthful description instead of recommending a tool or declaring one fixed legal status for all outputs.
Ethical or rights principle
AI tools can generate images, text, and layouts, but designers remain responsible for suitability, accuracy, and rights.
Outputs may contain logo-like material, similarities to known works, or incorrect text and therefore require human review.
Disclosure is needed when required by client or competition policy or when it materially changes understanding of the asset’s nature.
AI use does not remove the need for editable source files and a maintainable system after generation.
Sensitive client data should not be uploaded to external services without understanding privacy and use terms.
Before generation, distinguish a general task description from client files containing unpublished information. Easy uploading does not establish a service's suitability for every material. Relevant purpose and privacy terms should be examined as actually available, with authorization sought for what may be shared. Fictional names and figures can replace real data in the educational example. Their fictional status must remain in the project record so they do not later enter a final advertisement as supposed facts about a product or organization.
Responsible practice
Step 1: Review outputs for trademarks, text errors, and logical inconsistencies.
Step 2: Rebuild important assets in an editable design system rather than relying only on generated raster output.
Step 3: Record tools used when the project requires provenance.
Step 4: Review terms of service before uploading confidential data.
Output inspection has separate levels: does the image communicate the requested idea, are elements and relationships coherent, and are there words, marks, or forms resembling unintended assets? An attractive thumbnail is insufficient; problems may appear when reading a line or enlarging a region intended for print. Any writing should be checked against approved wording in both characters and meaning rather than trusted because it looks plausible. Absence of an obvious trademark likewise does not guarantee that every rights issue has been resolved.
Legal and professional context
UNESCO’s AI ethics recommendation emphasizes transparency and human oversight.
Rights issues around outputs and training data continue to evolve legally.
Computer-aided design has long used automation, while modern generation greatly expanded the volume of material produced quickly.
A tool illustration may show disconnected parts or repeated elements that alter the apparent use. The defective version can be rejected or a region rebuilt through a documented process, but correcting one noticed issue should not be described as comprehensive inspection. The revised version needs renewed review because one adjustment may affect another relationship. Technical assessment also differs from provenance review: a coherent image may raise rights questions, while permitted material may fail to serve the message. Neither track substitutes for the other.
Limits and risks
Generation speed does not equal identity consistency or system quality.
Accepting output without editable sources can create later operational burden.
The rights status of AI outputs should not be presented as one stable global rule.
Editability concerns future team needs. If wording, mark, and imagery are merged into one image, changing a date, size, or language may require rebuilding. Important components can be recreated as editable elements while retaining a reference version and documenting what remains raster material. A file bearing a design-program extension is insufficient when it contains only a flattened image. Delivery should describe actual maintenance capabilities rather than create expectations unsupported by the files.
Disclosure depends on purpose, policy, and audience understanding. Where a competition or client requires a tool statement, accurately describe use and subsequent team reconstruction or revision. Where a fabricated scene might be mistaken for photography of a real product or location, review how its nature should be clarified. Naming a tool does not make every pictured claim true or supply every required permission. Disclosure explains production, while factual and rights checks address separate questions that must not disappear behind it.
A provenance record can retain the task description, selected version, substantial changes, and external materials, limited to project needs without exposing private client data. Reuse in a new campaign requires renewed review of suitability, terms, and context rather than relying on approval for another purpose. Inspection limits should also be documented: do not claim complete originality or worldwide guaranteed rights without supporting evidence. A maintainable file, accurate statement, and identifiable checks are stronger professional outcomes than broad promises suggested by production speed or the quantity of generated alternatives.
Delivery quality can be tested through a simple requested revision, such as changing a date or language while preserving the visual system. If this requires regenerating the entire scene, the file should not be described as a flexible, maintainable component in the promised sense. State editing limits and choose an appropriate delivery method. Change records should also distinguish manual revisions from newly generated results, since provenance differences may affect reinspection even when versions share a filename.
Conclusion
Generated design is adopted after reviewing meaning, elements, provenance, and maintainability, with disclosure describing actual work and verification limits rather than treating generation speed as proof of originality or rights clearance.