Arts

After the Evidence

Tappin's Photograph Art Gallery, “[Union Cavalry Soldier with Pistol in Holster]”, (1861–65). The Metropolitan Museum of Art, Public Domain (CC0).

What photography can still prove when images can be generated without a camera

IE by intendle Editors Aug 7, 2026 · 11 min read

A photograph once carried a simple physical claim. Light from a scene entered a camera and left a trace on film or a sensor. That claim never made the image neutral. Photographers selected a frame, a moment, a lens, a position and an exposure. Editors cropped. Printers changed contrast. Governments staged scenes. Families removed people from albums. Artists used double exposure, collage, montage, darkroom manipulation and constructed tableaux almost from the beginning of the medium.

Yet something has changed. An image can now resemble a photograph without any photographed event having occurred. A face may belong to no person. A street may combine architectures that never stood together. A documentary-looking scene may have no camera position, no witness and no original moment to return to.

This does not end photography. It makes the medium's old tensions impossible to ignore.

Photography has always lived between evidence and invention. In 2026, its future depends on how clearly artists, journalists, museums and audiences distinguish those roles.

A photograph has never been the same as truth

The camera's mechanical reputation gave photography an authority that painting rarely possessed. A photograph seemed to say: this was there. But the history of photography is also a history of framing, omission, performance and interpretation.

Recent exhibitions have emphasized this constructed nature rather than treating it as a new digital crisis. The Norton Museum of Art's Blur / Obscure / Distort assembled works that use blur, concealment and visual disruption to unsettle the viewer's sense of time, space and scale. The exhibition's central point was not that photography suddenly became unreliable. It was that manipulation and perceptual instability have been part of photographic art since the nineteenth century.[1]

At SFMOMA, (Re)Constructing History examined how photographs can record an instant while carrying layers of historical power, appropriation and visual convention.[2] Leslie Hewitt's Achromatic Scales at the Norton combined found photographs, books, documents and studio arrangements to complicate the idea of photography as transparent observation.[3] The Nelson-Atkins Museum's More Is More revisited artists who multiplied, reconstructed and physically expanded photographs beyond the single frame.[4]

These practices matter because they prevent a false nostalgia. The challenge of synthetic imagery cannot be solved by pretending that analog photography was pure. The more useful distinction is not between manipulated and unmanipulated images. It is between accountable and unaccountable image-making.

Photography and the camera-made event

Photojournalism has responded to generative imagery by drawing a firm boundary. World Press Photo's 2026 rules state that AI-generated images are not photography because a photograph records light from a physical moment on film or a sensor. Synthetic images and generative fill are excluded. The rules also restrict content-changing manipulation and require entries to remain single-frame, single-exposure photographs.[5]

That definition is appropriate for a documentary contest. The public value of photojournalism depends on an actual event, a photographer's presence or camera-based record, and an auditable chain between capture and publication.

Art photography has a different mandate. An artist may use synthetic imagery, staged performance, digital compositing or algorithmic transformation as part of the work. The Isabella Stewart Gardner Museum's 2026 exhibition Persona traced photographic self-invention from face paint and alter egos to contemporary AI-assisted practices.[6] Such work is not fraudulent merely because it rejects documentary realism. Its integrity depends on whether the method is part of the artistic proposition and whether the audience is misled about how the image was made.

The same visual object can therefore be acceptable in one context and unacceptable in another. A fully synthetic image may be legitimate conceptual art. It cannot honestly function as camera evidence of a war, protest, crime or disaster that never occurred.

Context is not a footnote. It is part of the medium.

The collapse of visual innocence

The arrival of generative systems has accelerated three pressures that photography was already facing.

First, production has become abundant. A technically polished image can be generated in seconds, while a photographic project may require access, travel, trust, repeated visits, risk, editing and years of attention. Visual abundance can make this labor less visible.

Second, credibility has become expensive. Photographers and publishers must spend more time documenting provenance, retaining original files, managing metadata and explaining editorial decisions.

Third, audiences may respond to real photographs with generalized suspicion. The phrase "this could be AI" can become a way to dismiss inconvenient evidence without investigation.

This last danger is particularly serious. Synthetic imagery does not only create false pictures. It can weaken belief in authentic ones. The result is sometimes called the liar's dividend: once fabrication becomes plausible, people accused by genuine evidence can claim that the evidence is fabricated.

Photography therefore needs more than detection tools. It needs systems of trust that connect images to people, institutions and documented workflows.

What Content Credentials can and cannot do

The Coalition for Content Provenance and Authenticity, known as C2PA, has developed an open technical standard for recording the source and editing history of digital media. Version 2.4, released in April 2026, expanded format support, added new assertions and introduced a JSON-based representation for validation and interoperability.[7]

A Content Credential can contain digitally signed statements about creation, editing actions, software, capture devices and other provenance information. When an asset is altered, new information can be added while earlier provenance is retained.

This can answer valuable questions:

Which device or software created the file?
Was the image cropped, color-adjusted or composited?
Did an identified publisher sign the record?
Has the credential been altered or broken?
Is there a linked history of ingredients and edits?

It cannot answer every important question.

The C2PA specification is explicit that it does not judge whether provenance data is morally good, artistically valid or factually true. A valid signature can confirm that a known organization made a statement. It cannot prove that the caption is accurate, that a scene was not staged, that a photographer acted ethically or that an image represents the full context of an event.

Provenance is therefore infrastructure, not verdict.

A signed chain of edits may increase trust. It must still be combined with editorial standards, human verification, source protection, field reporting, caption accuracy and public accountability.

The artistic photograph after automatic realism

If machines can produce photographic realism without a camera, what remains specific to photographic art?

One answer is encounter. Photography can still be an act of being somewhere, waiting, negotiating access, observing change and accepting that the world will not obey the artist's prompt. The camera receives resistance from reality. Weather changes. A subject refuses. A gesture happens too early. A building blocks the view. Time exceeds the plan.

This resistance is not a guarantee of quality. It is a condition that distinguishes camera-based practice from fully generated imagery.

Another answer is duration. Anastasia Samoylova's Atlantic Coast retraced a historic route along the eastern United States, using documentary photography to examine landscape, ideology, consumerism and national myth.[8] Edward Burtynsky's The Great Acceleration assembled decades of work on industrial transformation and environmental damage.[9] Such projects are not valuable because a camera was used. They are valuable because observation accumulated into a sustained argument.

A third answer is the physical print. In an online feed, photographs are often reduced to interchangeable rectangles. The print restores decisions about scale, paper, surface, density, sequence, framing and exhibition. The continued vitality of photobook culture, reflected in the International Center of Photography's 2026 Photobook Fest with more than 80 publishers, shows that photographers still seek forms that cannot be reduced to a single scrolling image.[10]

The future of photography may therefore depend less on defending realism as a look and more on defending photography as a practice of attention.

The photograph as archive and evidence of process

Museums collecting contemporary photography now face a more complicated object. A photograph may exist as a raw capture, working file, layered edit, final master, print, social-media crop, video sequence and licensed derivative. Some versions may contain provenance metadata. Others may have been stripped by platforms.

A serious acquisition record should address at least five layers:

Capture: original raw files, device information, date, location where appropriate, and the circumstances of making.
Edit: a documented account of material changes, compositing, retouching, generative tools and software dependencies.
Context: caption, subject consent, commissioning conditions, relevant risks and the artist's intended mode of reading.
Edition: print size, paper, process, edition number, printer, color-management information and authorized variants.
Archive: file formats, checksums, migration plan, metadata, rights and instructions for future display.

This protocol should not become bureaucratic theater. Not every artwork requires the same documentation. Some artists deliberately work with ambiguity, anonymity or unstable files. But institutions should know whether that instability is intentional or accidental.

The museum cannot preserve only the visible image. It must preserve the conditions that allow the work to remain intelligible.

Manipulation, disclosure and artistic freedom

The language of "manipulation" often becomes morally confused. All photographs are processed. Even a camera's default image contains computational decisions about sharpening, color, noise reduction and dynamic range. Smartphone modes may combine multiple exposures or generate depth effects before the photographer sees the result.

The ethical question is not whether processing occurred. It is whether the processing changes the work's claim.

For documentary photography, removing a person or adding an object can falsify evidence. For a conceptual artwork, radical compositing may be the central method. For portraiture, skin retouching can become an issue of consent, body politics and commercial expectation. For archival work, colorization may help access while also risking historical misinterpretation.

Disclosure should therefore be specific. A label reading "AI-assisted" is often too vague to be useful. Did the artist use AI to generate the entire image, extend the background, remove an object, search an archive, create a mask, reduce noise or write a caption? These are not equivalent actions.

A professional disclosure can remain concise while distinguishing:

camera capture from synthetic generation,
documentary claim from artistic fiction,
local adjustment from content alteration,
human-authored elements from algorithmically generated ones,
verified provenance from unverified description.

Clarity does not diminish mystery. It prevents mystery from being used as cover for misrepresentation.

The future is not a purity test

Photography will not be saved by declaring one technology authentic and every later technology corrupt. The medium has changed with each new camera, printing process, distribution system and editing tool.

What it needs now is a stronger culture of declared method.

Artists should be free to construct impossible images. Journalists should be required to preserve the integrity of camera-based evidence. Museums should collect both the image and its technical history. Publishers should distinguish illustration from documentation. Audiences should learn to ask not simply "Is this real?" but "What claim is this image making, and what evidence supports that claim?"

A photograph can no longer rely on visual realism alone. Its authority must be earned through context, provenance and responsible authorship.

The medium's future begins after innocence, but it does not begin after truth.

Six questions readers are likely to ask

1. Is an AI-generated image a photograph?

It depends on the context in which the word is used, but a fully generated image is not camera photography in the documentary sense. World Press Photo defines a photograph as a record of light from a physical moment captured on film or a sensor, and excludes synthetic images from its contest. In contemporary art, an artist may deliberately create an image that looks photographic without a camera. That work can be valid digital art, but it should not be presented as evidence of a photographed event. Clear terminology protects both artistic freedom and documentary trust.

2. Can Content Credentials prove that a photograph is true?

No. Content Credentials can help verify a file's recorded provenance, including who signed information about it, what tools were used and whether the record appears intact. They do not prove that a caption is accurate, that a scene was not staged or that a publisher acted ethically. The C2PA standard deliberately avoids making value judgments. Provenance can strengthen verification, but it must be combined with reporting, editorial review, source checks and contextual evidence.

3. Does digital manipulation make photography less artistic or less authentic?

Not automatically. Photography has always involved choices and processing. In art, compositing, scene building, color alteration and synthetic elements may be essential to the concept. The relevant question is whether the method supports the work and whether the audience is misled about its claim. A documentary photograph has stricter obligations than a conceptual image. Authenticity does not require technical purity. It requires coherence between process, intention, presentation and context.

4. What should photographers disclose when they use generative tools?

They should disclose the kind of intervention that materially affects how the image is understood. Useful distinctions include full image generation, generative expansion, replacement or removal of content, synthetic subjects, AI-based retouching and routine enhancements that do not add new information. The disclosure should match the risk of misunderstanding. A museum label, editorial caption and commercial campaign may require different levels of detail, but "AI-assisted" alone is often insufficient.

5. Why do physical prints and photobooks still matter?

They give photography a material and sequential form that screens often erase. Paper, scale, surface, binding, image order and physical proximity influence meaning. A photobook can create an argument through sequence and pacing. A print can reveal tonal and textural decisions that disappear on a phone. Physical objects also provide more stable collecting and conservation frameworks, although they still require documentation and care.

6. How should museums preserve born-digital photography?

Museums should preserve more than the final visible file. Depending on the work, they may need raw captures, master files, color profiles, edit histories, software information, Content Credentials, checksums, edition records, display instructions and rights documentation. They should also identify which elements are essential and which may change over time. The goal is not to freeze every technical detail, but to preserve the artist's intended work and the evidence needed to reproduce or interpret it responsibly.

Sources and research notes

Comments

No comments yet.