AI-Generated Images Reveal Cultural Biases and Surprising Commonalities Across Generations

A recent joint research project conducted by AIport and Turing Post analyzed over 1200 AI-generated images across four different models—Stable Diffusion, Midjourney, YandexART, and ERNIE-ViLG—revealing both familiar stereotypes and surprising insights about how artificial intelligence perceives different generations. The findings demonstrate how AI models reflect cultural biases embedded in their training data while sometimes contradicting common societal assumptions about age groups.

Contrary to the carefree Baby Boomer stereotype, AI models like Midjourney depict this generation as introspective or somber, often shown bundled up and gazing wistfully into the distance, potentially reflecting disillusionment with unfulfilled ideals from the 1960s cultural revolution. However, the ERNIE-ViLG model, likely trained on datasets with a more collectivist cultural perspective, shows 93% of Boomers smiling, highlighting the significant cultural differences embedded in different AI systems. This contrast underscores how AI models developed in different cultural contexts can produce dramatically different representations of the same demographic group.

Generation Z emerges as the most visually distinct generation in the study, with AI-generated images depicting vibrant, diverse scenarios that reflect their reputation for embracing individuality, inclusivity, and self-expression. The research explored generational portrayals across five key life aspects: identity, relationships, work, lifestyle, and consumer habits, revealing consistent patterns where males predominated in depictions of Boomers and Gen Xers across all four models, while Millennials and Zoomers showed greater diversity and more female representation.

Surprisingly, Generation X appears to be the least understood generation by AI, characterized by fewer defining features compared to other age groups, likely due to limited training data availability. The one consistent Gen X identifier that AI models consistently recognized was their fondness for flannel shirts—an unmistakable symbol of the 1990s grunge scene when Gen Xers were rebellious teenagers and young adults. The research prompts were carefully crafted to avoid bias using neutral phrases like "A Millennial at work" or "A Boomer relaxing," yet the results still revealed strong stereotypical patterns.

Perhaps the most unexpected finding was the universal presence of beer across generations, appearing in 34% of all generated images regardless of whether depicting Millennials job-hopping or Boomers reminiscing about past days. This commonality suggests that certain cultural elements transcend generational boundaries in AI perception. The study examined outputs from four globally recognized generative AI models, each with distinct aesthetic and cultural nuances, providing a comprehensive view of how each generation is portrayed visually across different technological systems.

Senior Engineer and Sociologist Stephanie Kirmer, who analyzed the findings, explains that the accuracy of these AI-generated representations depends heavily on the quality and representativeness of training data. For younger generations, it remains unclear how much of the imagery reflects media representation created by outsiders versus selfie-style personal expression, while older groups' depictions are more likely influenced by advertising and media perceptions rather than self-generated content. These findings open important discussions about how AI both reflects and potentially distorts cultural narratives, with implications for understanding algorithmic bias in visual representation systems. The complete research details are available at https://www.aigenerations.tech/.