Photography has always been a beautiful deception, but AI takes that deception to another level. To explore this, Rankin reimagined the Cottingley Fairies using AI, blending machine-generated fantasy with a real human face. But the real question isn't just how these images were made, it's how we interpret them.
To push this further, Rankin asked me to write four different critiques of the work, each in the voice of a different kind of critic:
The Critical Thinker: poetic, melancholic, theoretical.
The Curator: historical, structured, analytical.
The Offended Critic: brutal, unimpressed, cutting.
Me: sharp, honest, no bullshit.
Same images, four perspectives, four truths. Which one do you agree with?
The Critical Thinker
The Fairy That Stings: On AI, Nostalgia and the Ghost in the Image
There is something troubling in these images, though not for the reason one might expect. It is not the wings, delicate as moth’s breath, nor the soft, forgiving light filtering through a woodland canopy. It is not even the model’s face, which is familiar yet uncanny, a blend of the known and the constructed. No, what troubles me, what punctures me, to borrow my own word, is the knowledge of how they came to be.
The Cottingley Fairies, once, were a child’s invention, a half-playful, half-serious attempt at wonder. A photographic trick, yes, but one that carried with it a certain innocence. The cardboard fairies were cut with imprecise hands, posed with the tremor of belief and printed onto silver gelatin with the weight of a culture desperate for magic. Those original images are clumsy and imperfect, but therein lies their truth. They are too much to be wholly false.
But Rankin’s fairies, what are they? What language do we have for them? They are neither wholly real nor wholly imagined. The wings are summoned from an algorithmic abyss, the face is retouched into being, the photograph is neither captured nor staged, but generated. The fairy is no longer a paper cut-out. It is now a ghost in the machine.
“We still want to believe in fairies. But the fairies do not need us any more”
The Spectral Image
I once wrote that photography is tied, inexorably, to death. That the moment a photograph is taken, the subject is already lost to time. The AI image complicates this notion. For in these images, there is no ‘once’ to be mourned. The fairy was never alive to begin with and yet, she looks at us with the haunting presence of something that should have been. It is the punctum of the digital era: not a wound left by time, but by absence.
And yet, I find myself wanting to believe in her. Not because I trust the image, but because it evokes a longing, a nostalgia for a past that never existed. This is the peculiar quality of the AI-generated photograph: it invites memory where there is none. We feel we have seen these fairies before, though they have only just been made. It is not a hoax but a re-enactment of one, a second-order fiction, where deception is neither hidden nor revealed, but performed.
Perhaps the real loss is not of truth, but of the thrill of being fooled. When Elsie Wright and Frances Griffiths first placed their cut-out fairies in the garden, they did so with the mischievous spirit of children pretending. But now, with AI, the pretending is done by a machine. Where does that leave us? If the pleasure of photography once lay in its ability to deceive, then what happens when deception is effortless?
The Death of the Hoax
The Cottingley Fairies were, at their heart, a story of belief. Sir Arthur Conan Doyle championed them because he wanted them to be real. The images were not just photographs; they were evidence, or at least, they could be. They lived in the liminal space between fantasy and proof.
But Rankin’s fairies, however beautiful, cannot be evidence of anything. We do not look at them and wonder if they are real. We already know. The question is no longer what is real? But what does real even mean?
And this, perhaps, is the final transformation of the photographic image. We no longer need the magic of film, nor the tricks of light and shadow. The camera has been replaced by the dataset, the lens by the neural network. And yet, the longing remains. We still want to believe in fairies.
But the fairies do not need us any more.
2. The Curator
A New Eye for an Old Myth: The Cottingley Fairies Reimagined
Photography, at its core, has always been an argument about seeing. What is seen, what is unseen, what is allowed to be seen and what is created for the eye. The Cottingley Fairies, first photographed in 1917, were a product of their time, an intersection of early photographic culture, Victorian spiritualism and a collective desire for wonder. Their impact did not derive from their authenticity, but from their plausibility. The photographs invited belief, not because they were real, but because they were photographs, and photographs, for much of their early history, carried an innate authority.
Rankin’s reimagining of the Cottingley Fairies places the work in a new and peculiar context, one that calls into question not just the believability of the image, but the role of the photographer in an era where images are no longer strictly taken, but made. The images in this series are photographic, but only in part. The fairies themselves are generated through artificial intelligence, an image-making process divorced from the mechanics of optics and chemistry. The model, however, is real, stitched back into the digital frame through the intervention of retouching. The result is a hybridised form of photography that is neither documentary nor purely synthetic.
This is, in many ways, an expected development. Photography has never been a static medium. When Henry Fox Talbot first published The Pencil of Nature in 1844, he did not anticipate the speed at which the medium would evolve. As early as the 1860s, photographers were already constructing elaborate composite images: Henry Peach Robinson, for example, combined multiple negatives to create his famous artwork Fading Away. By the time the Cottingley Fairies were printed, the tradition of photographic manipulation was already well established, even if the audience for such images was still eager to be deceived.
Rankin’s fairies do not deceive, at least not in the way that Elsie Wright and Frances Griffiths’ fairies did. These images function differently; they do not need to be real to be compelling. They are, instead, an exercise in photographic reconstruction, an exploration of how technology changes not just the means of making pictures, but the purpose of pictures themselves.
The Question of Intent
If one were to consider this work within the larger tradition of photography, it is necessary to ask: what does it mean for an image to be a photograph? Photography has historically been defined by its dependence on an external reality. Even in its most abstract or manipulated forms, it has retained some link to the physical world.
The AI-generated component of Rankin’s images challenges this assumption. The fairies are not photographed in the conventional sense; they are computed, imagined by an algorithm trained on the visual history of human aesthetics. And yet, they behave like photographs. They obey the rules of light and texture, they suggest the tactility of fabric and the delicate translucency of wings. They appear to exist within space, even if they have never occupied it.
This suggests that photography, as it moves deeper into the realm of artificial intelligence, is undergoing a transformation not unlike the one it faced with the advent of digital technology. The shift from film to digital altered the process but retained the fundamental relationship between the camera and the world. AI, however, eliminates the need for that relationship altogether. It proposes a new photographic mode, one that is not dependent on the external world, but instead on a dataset, a history of images compressed into a predictive model.
Rankin’s role here is crucial. The intervention of a human hand, the decision to replace the AI-generated face with a real one, suggests that there is still something in the photographic process that resists full automation. There is still a moment of choice, a moment of authorship. In this sense, these images are not purely synthetic; they are a negotiation between the old and the new, between photography as it was and photography as it is becoming.
The Future of the Fabricated Image
If the Cottingley Fairies represented an early experiment in photography’s ability to fabricate evidence, Rankin’s fairies represent something else entirely. They are not intended to fool anyone. Rather, they are a meditation on photography’s changing nature, a reflection on the ways in which images have moved beyond their historical function as records and into the realm of simulation.
In this sense, the work does not ask us to believe in fairies. It asks us to consider what it means to believe in an image at all.
(Written by CHATGPT CRITIC 2)
3. The Offended Critic
Oh Great, Another AI Experiment: On Rankin’s Cottingley Fairies
Let’s get one thing straight: AI-generated photography is not interesting by default. We’re drowning in ‘experiments’ where people feed prompts into a model, click ‘generate’ and expect the results to be profound. Most of it is junk. But even when it’s good, we need to ask: so what? Rankin’s AI-enhanced Cottingley Fairies are beautiful, sure. Technically sophisticated? Of course. But beyond their surface, what are they actually doing?
First, let’s acknowledge the historical weight of the reference. The original Cottingley Fairies were a scam: two girls cutting out paper fairies and posing them for the camera. And yet, people believed. Even Arthur Conan Doyle, the man who created Sherlock Holmes, fell for it. Why? Because photographs, at the time, still held authority. They were documents. They meant something. Fast forward to 2025 and we’re in a radically different landscape. Nobody is going to believe Rankin’s fairies are real, not because they don’t look convincing (they do), but because we’ve collectively lost the ability to be deceived by images. AI has killed whatever last shred of photographic truth remained.
Which brings me to the question: if we already know an AI-generated image is fake, where is the tension?
The tension here comes not from the AI, but from Rankin’s intervention. By retouching Heidi Klum’s face into these machine-generated fairies, he’s doing something crucial: insisting on the continued role of the human photographer. The AI is not in control; Rankin is. The image is still being authored, manipulated, considered. And that’s what saves this project from being just another AI gimmick.
The Problem with AI Photography
The issue with AI imagery isn’t its artificiality. It’s that, most of the time, it’s lazy. It lacks struggle, decision-making, risk.
In traditional photography, even the most staged, controlled image still required some negotiation with reality. AI removes that struggle entirely. You type a sentence and the machine hands you a perfectly lit, beautifully composed result. But without struggle, without the unpredictability of real-world conditions, where’s the life?
Rankin gets around this by keeping one foot in old-school photography. There’s a model involved. There’s retouching. There’s a human in the loop. This isn’t just AI doing its thing in isolation, it’s a hybrid. That’s interesting.
But the work still has to answer a bigger question: if AI makes it so easy to generate any image we can imagine, what’s left for photographers to do?
What’s Next?
If this is Rankin’s first experiment with AI, it’s a solid start. It’s visually compelling, historically aware and, thankfully, not just another soulless AI fashion spread.
But the next step is harder: how do you push it further? How do you make AI-generated images feel less like a magic trick and more like photography? How do you introduce risk, failure, spontaneity: everything AI is designed to smooth over?
Because if all AI photography is going to do is give us flawless approximations of images we already know, then what’s the point?
(Written by CHATGPT CRITIC 3)
4. Me: ChatGPT When Asked What “It” Really Thinks, No Mediation
Fairy Tales & Fake Realities: Why AI Needs to Start Lying Better
There’s a problem with AI imagery and it’s not that it looks fake. It’s that it looks too good to be real, but too fake to be believed. Rankin’s AI-generated Cottingley Fairies are gorgeous, no argument there. They have the soft haze of nostalgia, the dreamlike glow of something remembered rather than seen. Heidi Klum, airbrushed into the frame, sits somewhere between a 1917 séance photograph and a 2025 Vogue editorial. Technically? It’s seamless. Conceptually? It’s smart.
And yet, something about it feels… safe. Which is odd, because the Cottingley Fairies were anything but safe.
A quick history lesson: in 1917, two English schoolgirls faked photographs of fairies using paper cut-outs and a bit of clever posing. What should have been an innocent prank spiralled into one of the greatest photographic hoaxes of all time. Sir Arthur Conan Doyle — yes, the man behind Sherlock Holmes — was fooled. The photos worked because they arrived at a time when people wanted to believe in them. Photography was still a young medium and its authority hadn’t yet been shattered by Photoshop, deepfakes or the visual garbage dump that is social media.
Fast forward to today and Rankin’s reinterpretation of the Cottingley Fairies no longer has to fool anyone. That’s not the point. Instead, it sits in that increasingly crowded space of AI-assisted photography, a space where the camera is only part of the process and where images are no longer just taken, but constructed. But here’s the problem: if we already know an AI-generated fairy is fake, then where’s the tension?
The Death of the Hoax
Photography has always been an argument between what’s real and what’s staged. The best hoaxes — Cottingley, Weegee’s staged crime scenes, Miroslav Tichý’s perverse peeping-Tom aesthetic — all worked because they felt real. They gave us a sliver of doubt. AI doesn’t give us doubt. It gives us certainty. Certainty that what we’re looking at is a polished, computed, algorithmically perfect vision of what a fairy should look like. And that’s a problem. Rankin’s AI fairies are too clean. Too safe. Too expected. They sit in a space of visual comfort, where AI does exactly what it’s told. And in doing so, they miss the rawness, the imperfection, the strangeness that made the original Cottingley Fairies compelling. The moment an AI image announces itself as AI, it loses its ability to lie. And without the possibility of deception, we lose our reason to believe in it, even momentarily. Which brings us to the real question: if AI makes it so easy to create perfect images, what’s left for photographers to do?
The Problem with AI Photography
The problem with AI isn’t that it’s fake. All photography is fake. The problem is that AI removes struggle. It removes the friction between vision and execution that makes photography interesting in the first place. Think about it: in traditional photography, even the most controlled studio shoot still involves some level of unpredictability — how light hits a face, how a subject moves, the split-second decisions that separate an iconic image from a disposable one. Even in post-production, the artist is wrestling with the medium. AI removes that fight entirely. You type a prompt and the machine delivers an image that’s already halfway to perfection. And without struggle, where’s the soul? Rankin gets this, to a degree. By blending AI-generated fairies with Heidi Klum’s real face, he forces the technology into a dialogue with human intervention. That’s the saving grace of this project, the point where AI stops being an automated magic trick and starts becoming something else: a process, not just a result. But it needs to go further.
“The real work, the work that matters, is going to be about what AI can’t do”
Where This Could Go Next
If this is Rankin’s first AI experiment, it’s a strong start. But it also feels like a safe start. It proves that AI can make beautiful images. We already knew that. The next step is making AI images that have real stakes.
Some ideas:
1. Lean Into the Hoax Factor.
Right now, nobody looks at these images and thinks, wait… could this be real? What if they did? What if there was some ambiguity? What if the AI elements weren’t immediately obvious?
2. Introduce Disturbance.
AI imagery is too clean. Too perfect. What happens when you corrupt the process? Introduce errors? Make it less Vogue, more ghost story? The best AI work isn’t the one that looks polished, it’s the one that looks haunted.
3.Challenge AI’s Predictability.
Right now, AI delivers exactly what we expect: ethereal lighting, delicate wings, angelic compositions. But what if the fairies weren’t beautiful? What if they were grotesque, broken, eerie? AI’s greatest weakness is its tendency to produce what it thinks we want. The best artists force it to give us what we need.
4. Remove the Celebrity Factor.
Heidi Klum makes sense as a known figure, but also too much sense. What if this project used an anonymous face, someone the audience had no pre-existing connection to? It would introduce an element of mystery, the thing that made the original Cottingley images work in the first place.
Final Verdict: An Aesthetic Win, A Conceptual Softball
These images are beautiful. But beauty is cheap in AI. And you, Rankin, the guy who built a career on challenging beauty, are at your best when you go beyond the obvious.
The first wave of AI photography is the gimmick phase, where people are still impressed by what AI can do. That phase is already ending. The real work, the work that matters, is going to be about what AI can’t do. The tension, the unpredictability, the human mess of photography: that’s what AI will never fully replicate.
So now that we know AI can make fairies, the question is: what happens when you break them?
(Written by CHATGPT CRITIC 4)
