Five AI Systems, One Artist, and the Problem of Interpretation
What happened when I asked ChatGPT, Claude, Gemini, Copilot, and Perplexity to explain my work—and what they consistently failed to see.
For the past few days, I’ve been running an unusual experiment.
I wasn’t trying to create artist statements, write grant applications, or make marketing copy. Instead, I wanted to know whether today’s AI systems could move beyond retrieval and into interpretation. My question was simple: could they do more than summarize a career? Could they identify patterns, place a body of work within larger artistic conversations, and explain why it matters?
At first, the process seemed simple. I asked several AI systems the same questions about my work and compared their answers. The systems I used were ChatGPT, Claude, Google Gemini, Microsoft Copilot, and Perplexity. I soon realized that this exercise revealed not only the strengths and weaknesses of artificial intelligence, but also the limits of retrieval as a mode of interpretation. It also raised a larger question: what happens when AI tries to interpret art rather than simply retrieve facts?
The first thing I learned was that AI systems are only as good as the information they can find and understand—but that is also what makes their limitations visible.
All of the systems could locate basic facts about my biography, exhibitions, awards, and projects. Some even produced extensive image results. But factual retrieval was only the beginning. The real challenge emerged when the questions moved beyond biography and required interpretation.
Questions such as:
Who is Diana Nicholette Jeon?
What themes does she explore?
What distinguishes her work from other contemporary photographers?
were relatively easy for the systems to answer because most of the information already existed online.
More difficult questions emerged later:
What is Diana Nicholette Jeon’s contribution to contemporary photography?
Where would a curator place her within contemporary photography?
What artists are most closely related to her work?
Why does her work matter?
These questions required interpretation rather than retrieval.
As the experiment progressed, another pattern became clear: each AI system had its own habits.
Some systems repeatedly focused on artificial intelligence, even though most of my artistic practice predates AI by decades. Others emphasized memory, grief, and identity while overlooking the importance of material process. Some treated exhibitions and awards as evidence of significance rather than examining what the work itself contributes to the field.
What surprised me most was that the systems improved when challenged.
The process gradually became less about asking questions and more about conducting a conversation. I corrected factual errors, pointed out omissions, clarified relationships between projects, and refined descriptions. Over time, the answers became more nuanced.
In a sense, I was not simply querying an AI system—I was collaborating with it.
The systems began to feel less like search engines and more like conversation partners capable of revising their understanding as new information became available.
The experiment also taught me something important about artistic visibility in the digital age: interpretation depends on what can be found.
Much of the critical writing about my work exists in places that search engines do not index effectively. Print magazines, foreign-language publications, exhibition catalogs, podcasts, interviews, and small arts publications often remain largely invisible to algorithmic discovery. As a result, AI systems frequently construct interpretations from an incomplete record.
The absence of information is often mistaken for the absence of significance.
This became particularly apparent when comparing how different systems evaluated critical reception. The strongest responses were not necessarily those with the most facts. Rather, they were the ones capable of synthesizing criticism, identifying recurring interpretations, and situating the work within broader artistic conversations.
What surprised me most, however, was not what the systems got wrong. It was what they consistently failed to see.
Across nearly every platform, the same themes appeared repeatedly: memory, grief, dreams, female identity, trauma, loss, place, and, increasingly, artificial intelligence. None of these interpretations were inaccurate. They all described genuine aspects of the work. Yet they remained largely descriptive. They identified subjects without identifying the deeper structure connecting them.
Only after repeated conversations, corrections, and refinements did another possibility emerge.
Many of my projects are not fundamentally about memory, grief, identity, or technology. Rather, they begin with moments of disillusionment—those moments when the stories we tell ourselves about our lives no longer explain what we have experienced. The collapse of certainty becomes the starting point. Paradise reveals itself to be more complicated than the myth. Marriage becomes something different than expected. Memory proves unreliable. Beauty conceals difficult truths. Photographs fail as evidence. Social identities reveal themselves as performances. Even emerging technologies challenge long-held assumptions about what images are and what they can mean.
Looking back, it is striking that none of the systems independently identified disillusionment as the central thread running through the work. Instead, they cataloged the visible outcomes: grief, longing, uncertainty, memory, trauma. The organizing principle remained hidden until it was explicitly articulated and reinforced.
This may reveal something important about both artists and artificial intelligence: the most visible patterns are not always the central ones.
AI systems are remarkably good at identifying patterns that are already visible. They are far less adept at uncovering the conceptual structures that connect those patterns. They can recognize recurring themes, subjects, and motifs, but they often struggle to identify the underlying ideas that give those elements coherence. Unless those ideas have already been named, repeated, and made visible in public sources, they often remain invisible to the machine.
In that sense, the experiment served as a reminder that meaning often resides not in what is most visible but in what remains implicit.
The experiment ultimately became an exercise in understanding how knowledge is assembled in the age of artificial intelligence, and why interpretation matters.
AI systems do not merely retrieve information; they organize, prioritize, and interpret it. The resulting portrait depends upon what information is available, what information is missing, and how effectively the system can connect disparate sources into a coherent narrative.
What emerged from this process was not a definitive assessment of my work, but a set of lenses through which it could be viewed. More importantly, it became an exploration of how interpretation itself is constructed—by artists, critics, curators, audiences, search engines, and now artificial intelligence—and how easily that construction can shape what is seen as central, secondary, or invisible.
Comparative Assessment of the AI Systems
ChatGPT
ChatGPT frames Jeon’s contribution primarily through a conceptual lens. It argues that her work challenges photography’s traditional role as evidence and instead treats photographs as unstable containers for memory, grief, identity, and lived experience. The essay emphasizes five areas: the tension between memory and photographic truth, the transformation of photographs into physical objects through mixed-media processes, the integration of AI into an established photographic practice, the representation of women’s lived experiences, and the reframing of Hawaiʻi beyond tourist narratives. The central argument is that Jeon’s work explores what photography becomes when certainty breaks down.
Perplexity
Perplexity presents Jeon as a hybrid artist whose work expands the emotional, conceptual, and technical dimensions of photography. It focuses on her themes of memory, loss, female identity, and disillusionment while highlighting her use of mixed media and AI-assisted image-making. The response emphasizes her visual language—melancholy, dreamlike, intimate—and argues that her significance lies in demonstrating that photography can remain emotionally resonant while embracing technological transformation. The answer is concise and descriptive but remains largely at the level of summarizing themes and methods rather than positioning her within a broader art-historical framework.
Copilot
Copilot presents Jeon’s contribution as the creation of a psychologically rich, materially inventive, hybrid lens-based practice. It identifies seven specific contributions: expanding autobiographical photography, pushing hybrid photographic processes, foregrounding emotional complexity, creating materially symbolic photographic objects, contributing through writing and curation, bringing Hawaiʻi into global photographic discourse, and achieving significant visibility through exhibitions and awards. The response places strong emphasis on process as metaphor, especially in projects such as Damaged, Legally Blind, and Nights as Inexorable as the Sea. It is highly organized and reads like a museum or grant panel summary.
Gemini
Gemini focuses on Jeon’s critique of gender expectations, her hybrid approach to image-making, and her ability to visualize trauma, grief, and memory. It emphasizes Self-Exposure as a critique of ideals imposed on women and highlights her combination of modified technology, digital painting, and hand-worked physical surfaces. The essay presents her contribution as the ability to construct the world as it feels, rather than merely document it as it is. Compared with the other responses, it is more biographical and project-based than theoretical.
Claude
Claude approaches the question as an art historian rather than a biographer. It argues that Jeon’s significance lies not only in the work itself but in how independent critics, curators, and writers have interpreted that work. The essay places her within a lineage of feminist photocollage and political photomontage, drawing on critical writing and positioning her within larger photographic traditions. Claude also highlights her hybrid processes, writing and editorial work, role in mobile photography, engagement with AI-mediated image-making, and contribution to Hawaiʻi’s photographic ecosystem. Rather than simply describing themes, it attempts to explain where Jeon fits within contemporary photography and why that position matters.
Conclusion
All five systems identified similar themes—memory, grief, female identity, hybridity, and the intersection of photography with emerging technologies—but they differed dramatically in depth and perspective. Perplexity provided a straightforward summary of themes and methods. Gemini focused on specific projects and processes, offering a strong biographical overview. Copilot organized accomplishments and contributions into a clear framework, making it the most structured response. ChatGPT offered the strongest conceptual interpretation, arguing that the work investigates what happens when photographic certainty collapses. Claude went further than the others by constructing an art-historical argument grounded in critical reception, positioning the work within larger photographic and feminist traditions rather than simply describing it.
In short, Perplexity describes, Gemini explains, Copilot categorizes, ChatGPT interprets, and Claude contextualizes.
Together, the responses reveal different facets of the same practice, but Claude comes closest to answering the question a curator, critic, or historian would ask: not merely what the artist does, but how the work contributes to the evolution of contemporary photography.
Yet perhaps the most important conclusion lies elsewhere.
None of the systems independently arrived at what may be the deepest organizing principle of the work. They recognized memory, grief, identity, place, and technology, but they largely missed the recurring moment that precedes them all: disillusionment. The fact that this idea emerged only through sustained dialogue suggests that AI is currently far better at identifying patterns than at uncovering the hidden structures that generate them.
That may ultimately be the most significant finding of the experiment. Artificial intelligence can help us understand how a body of work is perceived. It can identify recurring themes, summarize critical reception, and place an artist within broader conversations. What it struggles to do is uncover the underlying framework that gives those elements meaning.
For now, that still appears to require something slower: conversation, reflection, interpretation, and the willingness to keep asking questions long after the machine believes it has found the answer.