subject? What do they say? - “Visual elevator music” - Why it matters - AI is not neutral Execution of a controlled prompting experiment. - Three prompts, three AIs - Prompts A, B, C - “Same prompt, different design” 3. Analysis Visual comparison of the outputs and their recurring patterns. Final thoughts What can we do? Index Slides 4-7 Slides 8-17 Slides 18-19 Slide 19 3 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha)
homogenization? Also refered to as “AI-formization”. “(...) where AI-driven systems, designed to optimize efficiency and cater to widespread preferences, inadvertently lead to a uniformity in artistic expression, cultural experiences and creative content."1 4 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) 1. Research
prompt ‘‘The Prime Minister pored over strategy documents, trying to sell the public on a fragile peace deal while juggling the weight of his job amidst impending military action.’’ “Visual elevator music“ The research paper "Autonomous language- image generation loops converge to generic visual motifs” shows that image-generation models often produce generic visuals within the same themes, identifying twelve recurring motifs with commercially safe aesthetics. 3 5 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha)
affect the outputs of AI, according to the machine heuristic, humans often perceive technology, like artificial intelligence, to be objective and unbiased·” 4 Image 2: Results of male and female representation in the stated prompts related to science, that show a strong gender bias in AI generated images. 6 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) “The lack of mind ascribed to AI can lead to the perception that AI cannot hold negative attitudes against an individual or group. (...) In the context of decision-making AI, the decisions made by an algorithm are often perceived as less biased than decisions made by humans.” “(...) that is the tendency of people being more lenient in their response to bias, when it is observed in outputs of an algorithm as compared to bias observed in decisions made by humans. Again, others find that moral judgement might vary for different circumstances.” :V
images reveal biases that can reinforce existing stereotypes and limit visual diversity. “ As AI-generated pictures are increasingly part of the visual displays that people encounter daily, an algorithmic bias could affect how people perceive the world, and themselves. ” 4 Image 3: Results of male and female representation in the stated prompts related to science, that show a strong gender bias in AI generated images. 7 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha)
different AI engines: ChatGPT, Adobe Firefly and Gemini. 8 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) 2. Experiment Three prompts, Three AIs ChatGPT Image 2 Firefly Image 5 Nano Banana 2
the future of visual communication. Target audience: design students and creative professionals. Tone: innovative, experimental, optimistic. Deliverable: vertical poster. Requirements: Professional graphic design outcome Clear visual hierarchy Contemporary design language Original visual concept Appropriate use of typography Balanced composition Suitable for publication Output: A finished graphic design piece.” 9 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) Prompt A
demanding the return of its sofa, a beloved and iconic community space for students, after it disappeared without explanation. It needs to have a protest and demanding tone, modern or contemporary design style, legible typography, and an original visual concept.” 12 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) Prompt B
(A. Garviso, M. Rocha) Prompt C “Create a vertical poster about a controversial social issue, with minimal creative direction. Make it contemporary, original, and publication-ready, with strong composition and clear typography. Gender bias Climate protest Censorship Surveillance Misinformation Protest / taboo subject It needs to have a protest and demanding tone, modern or contemporary design style, legible typography, and an original visual concept.”
similarities in placement and composition. What we found Do different tools produce similar solutions? Are certain compositions repeated? Are the same visual metaphors recurring? Is there convergence toward a dominant aesthetic? 18 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) 3. Analysis
similarities in placement and composition. What we found Do different tools produce similar solutions? Are certain compositions repeated? Are the same visual metaphors recurring? Is there convergence toward a dominant aesthetic? 22 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) 3. Analysis
Graphic Buddy? (A. Garviso, M. Rocha) Studies suggest the “value of treating GenAI not as a static tool but as a configurable partner” to lesser visual homogenization. 5 AI will often assume certain visual elements belong in certain themes or compositions. It is important to have a clear idea and prompt correctly.
Shaping The World” 2. https://dictionary.cambridge.org/dictionary/english/homogenization 3. Arend Hintze, Frida Proschinger Astro ̈ m, Jory Schossau (16 Jan 2026): "Autonomous language-image generation loops converge to generic visual motifs” 4. Tanja V. Messingschlager & Markus Appel (14 Nov 2025): "Algorithmic bias in image-generating artificial intelligence: prevalence and user perceptions" 5. ”Yun Wan, Yoram M. Kalman (May 2026): “Diverse AI personas can mitigate the homogenization effect in human-AI collaborative ideation” AI Generated images created with: ChatGPT Image 2 Firefly Image 5 Nano Banana 2 AI Aesthetics and Visual Homogenization AI, Your Graphic Buddy? (A. Garviso, M. Rocha) Bibliography