Absentia: AI, the Gendered Gaze, and Photography’s Unequal Encounter (Essay)
The asymmetrical power dynamics involved in photographic portraiture have always given me pause. Photography begins with a social encounter organized around a basic inequality: someone holds the camera and someone else stands before it. The photographer decides when to release the shutter, what to include, what to exclude, which image survives, and where it will circulate. The subject may collaborate, resist, negotiate, or refuse, but the photographer generally retains greater control over the construction of representation.
For much of photography’s history, this imbalance has also been gendered. Portraiture, fashion, advertising, and the nude developed within cultures in which men disproportionately occupied positions of creative and institutional authority while women were disproportionately made visible. John Berger’s Ways of Seeing condensed this history into a familiar proposition: men act and women appear. Laura Mulvey later gave this structure a more explicitly psychoanalytic formulation in “Visual Pleasure and Narrative Cinema,” describing a visual order in which looking is coded as active and masculine while being looked at is coded as passive and feminine.
Mulvey was writing about cinema, but the implications for photography are clear. A camera is not innocent. It does not simply record a body; it helps organize a relationship between the person authorized to look and the person made available to be seen. That relationship becomes especially consequential when the photographer also controls selection, editing, publication, and circulation.
This does not mean that every photograph made by a man of a woman is exploitative, or that women have been passive participants in photographic culture. Such a claim would flatten a complicated history. It means that consent operates within larger structures of authorship, economic power, institutional authority, professional ambition, and access. A subject can willingly agree to be photographed while still having limited control over which image is chosen or what meanings are attached to it after it enters circulation.
Susan Sontag recognized something acquisitive in this structure, repeatedly connecting photography with taking possession of appearances. Ariella Azoulay later complicated the idea of the photographer as sole author by describing photography as a set of relations among photographer, photographed person, and spectator. In both accounts, a photograph of a person is more than an image. It also holds the residue of a human encounter.
Generative AI introduces a significant rupture in this history because it separates the appearance of a person from the interpersonal encounter traditionally required to produce that appearance. This does not make synthetic imagery inherently more ethical. It does, however, allow us to distinguish two forms of power that photography has often bundled together.
The first is representational power: the authority to determine how gender, sexuality, race, class, age, beauty, vulnerability, or difference are pictured. The second is embodied power: the authority exercised during production, when one person directs another person’s body, expression, clothing, behavior, or performance in pursuit of an image.
AI does little to solve the first problem. In fact, it can reproduce existing visual prejudices with remarkable efficiency. Research on text-to-image systems has documented persistent gender stereotyping in generated imagery, including associations between masculinity and authority and between femininity and caregiving, beauty, or supportive roles (AlDahoul, Rahwan, and Zaki 2025). The male gaze can survive perfectly well as a visual convention even when no male photographer is physically present. Bias can migrate from the studio into datasets, model architectures, prompting conventions, and the visual archive on which generative systems depend.
The second form of power is different. It concerns not what an image means, but what had to happen between people for the image to be made. This is where generative image-making creates an ethical possibility that deserves more attention. If representation can be produced without directing an actual person’s body, then one particular site of coercion, negotiation, and vulnerability can be removed from the production process.
Sara Oscar and Cherine Fahd address the disappearance of the photographic encounter in their 2026 essay “Looking at Eyes That Never Looked Back.” They argue that synthetic portraiture severs an important relational structure of photography because the image no longer records a reciprocal encounter between photographer and subject. Their concern is persuasive. Something is lost when portraiture no longer carries evidence that one person stood before another and returned the camera’s gaze.
But that loss should not be understood only negatively. The photographic encounter has never been ethically neutral. It can be collaborative, respectful, even intimate, but it can also be shaped by unequal status, money, gender, reputation, or professional dependency. At its worst, it can place a subject’s body into a position of potential servitude to someone else’s image. If the encounter is one place where these differences become operative, then its disappearance can represent an ethical gain as well as an ontological loss.
The history of the photographed female body makes this distinction especially visible. Consider the familiar structure of a culturally powerful male photographer producing sexualized images of a female model. Two different ethical questions arise. The first concerns representation: What does the image say about women, desire, beauty, or availability? The second concerns production: What did the person being photographed have to negotiate, tolerate, or risk in order for that image to exist?
Those questions are related, but they are not the same. A synthetic image can still objectify women, reproduce narrow standards of beauty, or rehearse patriarchal fantasies. It can therefore fail completely at the level of representation. Yet the ethical conditions of production are no longer identical to those of a conventional photo session. The harm embedded in the image and the harm that might occur in making it have become separable.
That separation matters because photographic criticism has often concentrated on the image as representation while giving less attention to the interpersonal conditions through which that representation was produced. The photograph can be analyzed for ideology, symbolism, beauty, violence, race, gender, or class, but the studio encounter itself can disappear from view. Generative imagery makes that hidden layer easier to see precisely because it demonstrates that the two forms of power are not inseparable.
This argument has obvious limits. It applies only to genuinely synthetic subjects. Deepfakes and other images that appropriate the likeness of identifiable people can intensify violations of consent rather than diminish them. Training data presents another problem. Generative systems are built from vast visual archives whose histories of authorship, labor, bias, and exploitation remain embedded within them. Synthetic image-making does not emerge outside culture simply because its immediate subject is fictional.
Nor does AI erase the photographer’s gaze. It relocates it.
This relocation may be the more consequential point. In conventional photography, power has often been concentrated in the encounter among photographer, subject, camera, and institution. In generative image-making, more of that power shifts toward the archive, the model, the prompt, the platform, and the artist’s decisions about what kinds of bodies and identities to construct. The ethical burden moves from managing a live interpersonal relationship toward understanding the histories and assumptions embedded in the representational system itself.
Recent writing on AI-generated photography has understandably focused on what is lost when indexicality and physical presence disappear. Amanda Wasielewski has examined how synthetic photographic images complicate photography’s relation to reality, while Avery Slater has considered what happens to the gaze when the apparent subject of a photograph has no reciprocal relation to a photographer. These are central questions for post-optical photography, but they also open another one: what forms of power disappear, persist, or intensify when the photographic encounter is no longer the primary site of image production?
The answer is not that AI offers a morally superior form of photography. It does not. The more useful claim is narrower. Generative image-making separates representational ethics from the ethics of the photographic encounter in a way that traditional photography generally could not. That separation allows us to reconsider where responsibility resides.
For the photographer working with another person, ethics begins partly in relation: How much authority do I have here? What am I asking of this person? What pressures, explicit or implicit, shape their ability to refuse? For the artist constructing a synthetic subject, the emphasis shifts toward another set of questions: What visual histories am I reproducing? What assumptions are built into the body I am constructing? What cultural work will this image perform once it circulates?
Neither set of questions is easier than the other. They simply describe different structures of responsibility.
This is why the emergence of synthetic imagery should matter to photography theory beyond debates about whether AI images are “really” photographs. The more significant issue may be that post-optical image-making reorganizes the social relations historically attached to photographic production. The gaze remains. Representation remains political. The archive remains implicated. But the distribution of power changes.
Photography has spent nearly two centuries teaching us to think about the ethics of looking at other people. Synthetic image-making now asks us to consider the ethics of constructing people who exist only as images. Between those two conditions lies a meaningful shift: not the end of the gaze, but a change in where its power is exercised and where its consequences must be confronted.
References
AlDahoul, Nouar, Talal Rahwan, and Yasir Zaki. 2025. “AI-Generated Faces Influence Gender Stereotypes and Racial Homogenization.” Scientific Reports 15: 14449.
Azoulay, Ariella. 2008. The Civil Contract of Photography. New York: Zone Books.
Berger, John. 1972. Ways of Seeing. London: Penguin Books.
Mulvey, Laura. 1975. “Visual Pleasure and Narrative Cinema.” Screen 16 (3): 6–18.
Oscar, Sara, and Cherine Fahd. 2026. “Looking at Eyes That Never Looked Back: Photographic Portraits, Synthetic Images and (Mis)recognition.” photographies 19 (1): 145–166.
Slater, Avery. 2025. “Critical Response IV: This Photo Does Not Exist: Generativity and the AI Gaze.” Critical Inquiry 51 (2): 416–422.
Sontag, Susan. 1977. On Photography. New York: Farrar, Straus and Giroux.
Wasielewski, Amanda. 2024. “Unnatural Images: On AI-Generated Photographs.” Critical Inquiry 51 (1):