Graphic designers and AI
A professional-design field experiment found better ideation creativity alongside longer task time; implementation-only AI use produced no statistically significant improvement in measured creativity.
What published evidence shows
Graphic designers create visual concepts and layouts, confer with clients, and check final work.
Scope: US official occupation description
No measured AI capability finding.
Study 2 randomly assigned 120 employed graphic/visual design professionals to four AI-access conditions in an advertisement competition. It used Midjourney V6.1, expert creativity ratings and recorded stage time.
Scope: One external design competition; no empirical transfer to all occupation tasks
One advertisement competition; task-specific creativity and time outcomes, not occupation-wide productivity.
Study 2 used Midjourney V6.1. Student expertise groups in Study 1 are separate from the employed professionals in Study 2.
No significant creativity improvement is not proof of equal quality. No employment-displacement inference.
AI use during ideation improved rated novelty, relevance and complexity but increased time. Implementation-only AI use produced no significant creativity improvement and required more time than control.
Scope: Study 2 stage-specific randomized comparisons
Benefits differ by stage; higher creativity ratings do not establish faster work or accepted client-ready production.
Task boundaries
These are reference tasks and proposed evidence links, not verified AI capability ratings.
- Create designs, concepts, and sample layouts, based on knowledge of layout principles and esthetic design concepts.
onet:27-1024.00:299 · Untested occupational context. Untested occupation responsibility; contextual reference only. No empirical task transfer has been established.
- Use computer software to generate new images.
onet:27-1024.00:301 · Untested occupational context. Untested occupation responsibility; contextual reference only. No empirical task transfer has been established.
- Confer with clients to discuss and determine layout design.
onet:27-1024.00:305 · Untested occupational context. Untested occupation responsibility; contextual reference only. No empirical task transfer has been established.
- Review final layouts and suggest improvements, as needed.
onet:27-1024.00:304 · Untested occupational context. Untested occupation responsibility; contextual reference only. No empirical task transfer has been established.
What remains unknown
- Occupation-wide autonomous capability
- Representative employer adoption rate
- Net measured human effort in the Kingy setting
- Causal AI employment displacement
- Personal replacement probability
Questions for a real workflow
Compare ideation and production separately. Record all time, rejected images and revisions; assess brief fit, typography, licensed assets, production constraints and final client acceptance.