For more than a century, the Rorschach inkblot test has provided insight into human personality and perception. Even if one has never taken the test, its iconic mirrored ink smudges are instantly recognizable. Developed in 1921 by Swiss psychiatrist Hermann Rorschach, the test presents ambiguous inkblot images to individuals, asking them to interpret what they see. The interpretations vary widely and can reflect personal experiences, emotions, and cognitive biases.
The test operates on the principle of pareidolia, a psychological tendency to find familiar patterns in random stimuli—similar to seeing shapes in clouds or faces on the Moon. While many psychologists today argue that the Rorschach test is outdated and unreliable as a psychometric tool, it is still used in some therapeutic contexts and, controversially, even in legal cases. Rorschach originally designed the test to identify disordered thinking in patients, particularly those with schizophrenia.
According to Barbara Santini, a London-based psychologist who continues to use the test, "When a person interprets a Rorschach image, they unconsciously project elements of their psyche such as fears, desires, and cognitive biases." She emphasizes that human vision is an active, meaning-making process shaped by personal experience.
Recent advancements in artificial intelligence have allowed researchers to explore how AI models, such as OpenAI’s ChatGPT, interpret Rorschach inkblots. Unlike humans, AI lacks personal experiences and emotions. However, multimodal AI models—those capable of processing both text and images—can analyze visual inputs and generate interpretations.
When presented with a common Rorschach inkblot, ChatGPT initially acknowledged its ambiguous nature before offering a structured analysis: "For me, it resembles something symmetrical, possibly two animals or figures facing each other, or a single entity with wings outstretched." When pressed for a more definitive answer, it settled on "a single entity with wings outstretched—perhaps a bat or a moth."
Dutch software developer Coen Dekker, who experimented with an earlier neural network to complete a Rorschach test, suggests that AI’s responses are shaped by vast datasets rather than personal perception. Similarly, psychologist Ieva Kubiliute notes that AI does not truly "see" images but rather recognizes patterns and textures, then generates interpretations based on existing human responses.
Psychologist Chandril Ghosh from the University of Kent explains, "ChatGPT provides interesting, human-like responses, but it is likely not genuinely thinking—it appears to be scanning an online dataset." He compares it to a musician who has never experienced heartbreak but can compose an emotional song by analyzing the structure of existing music.
The way AI generates interpretations of inkblots highlights a fundamental difference between human and machine cognition. AI lacks true subjectivity and instead mirrors collective human visual culture. This becomes evident when the same inkblot is presented multiple times—ChatGPT may provide different interpretations in each instance.
"A human would typically stick to their previous answer because personal experiences and emotions influence their responses," Ghosh explains. "In contrast, ChatGPT generates responses based on its dataset, meaning its answer can change with each request."
A striking example of how AI training data shapes perception is an experiment conducted at the Massachusetts Institute of Technology (MIT). Researchers trained an AI model, dubbed "Norman," on violent and gruesome images from a dark online forum. When Norman was shown a Rorschach inkblot, it described seeing disturbing and violent imagery, whereas a standard AI model saw birds or tree branches. This underscores how AI’s "thoughts" are entirely dependent on its training data.
While AI algorithms excel at recognizing patterns, their responses lack emotional depth and subjective meaning. AI cannot attach symbolic significance or emotional resonance to an image in the same way humans do. "The human psyche is filled with internal conflicts, such as the tension between desires and morals or fears and ambitions," says Ghosh. "In contrast, AI functions on clear logic and does not struggle with inner dilemmas essential to human thought and decision-making."
This distinction reveals a fundamental truth: while AI can simulate perception, it does not truly "see" or interpret in a human sense. The Rorschach test, originally designed to probe the depths of human thought, remains an imperfect but uniquely human exercise—one that AI, no matter how advanced, can only mimic without true understanding.