What is real in photography?

Alan Hodgson’s article, ‘What is real and what is not?’, in Issue 7 of the RPS Imaging Science Group newsletter (October 2026), gave me plenty to think about. I agree that treating photography as real and generative AI as unreal leaves too much unexplained. My own feeling is that we need to ask what we mean by ‘real’ before deciding which images deserve the word.

For me, three meanings are especially useful: the physical reality of the subject, the reality of our visual experience, and the reality of the camera’s recording. Keeping them separate helps explain why a photograph can be authentic without resembling ordinary sight, and why an AI image can look convincing without documenting an actual event. Artistic intention adds another question, concerning the creator’s relationship with the image.

From film to digital

The present debate reminds me of the transition from film to digital photography. There was an understandable attraction to the idea of a negative as a physical trace of something that had stood before the camera. Digital files seemed more open to alteration. With Photoshop, colour and contrast could change, saturation could increase, and objects could disappear or arrive. If the picture was so easy to alter, how much trust should we place in it?

That concern has a familiar sound today. However, as Alan’s historical examples remind us, pre-digital photography was also open to manipulation. Retouching, multiple exposures, combination printing and photomontage existed long before Photoshop, as the Metropolitan Museum of Art’s exhibition ‘Faking It’ documented.⁠¹ Digital tools made many alterations easier and more accessible. They did not introduce the possibility of changing a photograph.

The analogy is useful, although it has a limit. Film and digital cameras both begin by recording light. A wholly generated image need not begin with an exposure of the subject it depicts. That difference matters whenever an image is offered as evidence. Still, before considering AI, I think it is worth looking at the ways ordinary photography already departs from our experience of a scene.

Colour and the experience of seeing

Early photography was largely monochrome, often on plates rather than the film we now have in mind. Photography’s public introduction in 1839 preceded the commercial arrival of the Lumière brothers’ Autochrome colour plates in 1907 by nearly seven decades.⁠² Kodak introduced Kodachrome in 1935, initially for 16 mm motion pictures, with 35 mm still photography following in 1936.⁠³

That history raises a question. If a monochrome photograph of a daylight landscape omits the colours we saw, in what sense does it show the scene as it really appeared? The grass, flowers and sky did not lose their colours because the photographer used a particular material. We accept the photograph as a translation into tones, and we may find that translation more expressive than a colour image.

At night, the comparison becomes more interesting. Our rods are particularly sensitive in low light and do not provide colour discrimination on their own. Cones support colour vision when there is sufficient light.⁠⁴ In very dim conditions, our experience can therefore become largely monochromatic. However, rods and cones can both contribute at intermediate light levels. Vision does not simply switch from black and white to colour at one universal threshold.⁠⁵

Nor does this make a black-and-white photograph an exact reproduction of night vision. Photographic materials and our eyes respond differently to wavelengths and brightness.⁠⁶ A monochrome photograph may resemble one aspect of our experience without reproducing the whole process. We can recognise the landscape in either version while acknowledging that both have translated it.

Colour photography offers another translation. Film’s response and processing, and the conditions in which we view the result, all influence the appearance.⁠⁶ Digital photography has its own conversion from measurements to a visible image. A RAW file gives us considerable freedom to choose a rendering; while a JPEG already incorporates processing decisions.⁠⁷ Colour reproduction can be carefully controlled, but calling a photograph ‘colour’ does not establish that it matches everyone’s visual experience.⁠⁶

Our eyes also involve conversion. Light triggers biochemical changes in photoreceptors, altering their electrical activity.⁠⁸ The retina processes the resulting information before signals travel through the optic nerve to the brain.⁠⁹ Vision also varies from person to person. Adaptation to the surrounding light and the context in which we see a colour can affect our experience.⁠⁶ ‘What I saw’ is a useful account, but it is already an account from a particular observer.

The journey of light

There is also the journey of light before it reaches either an eye or a camera. Gravitational lensing provides an extraordinary example. A massive object can curve spacetime so that light from something behind it follows a bent path. The distant object may appear displaced, magnified, distorted or in multiple images.⁠¹⁰ A telescope can accurately record those effects without showing the object as it would appear along an unaffected line of sight. The distortion is a real consequence of the light’s journey. It shows why the physical subject and its observed appearance should be distinguished. What we receive is information that has travelled through particular conditions. Those conditions help determine the image, even before we choose a lens or touch an editing control.

Sunrise and sunset bring this much closer to home. The Sun can appear orange or red because its light travels through a longer path in the atmosphere. Shorter wavelengths are scattered out of the direct beam more strongly, leaving the light reaching us relatively richer in reds and oranges.⁠¹¹

The Sun emits a broad visible spectrum which, taken together, appears approximately white without that atmospheric filtering.⁠¹¹ Yet the red appearance is real for the observer in those circumstances. The Sun has not transformed into a different kind of star, but the light arriving here has changed. A photograph of the red Sun can accurately show that appearance.

Focus and time

A camera introduces further conditions. We often choose a large aperture to soften a background and draw attention to a person or flower. With other factors held constant, a larger aperture gives a shallower depth of field. The background objects remain present and detailed, but their light forms an out-of-focus image at the recording surface. The softness comes from the optical arrangement.⁠¹²

Our eyes also have optics and limited depth of field. As we look around, we shift our attention and refocus. A conventional single exposure usually records one chosen focus arrangement. Focus stacking can combine exposures made at different focus positions to extend the depth of field.⁠¹³ A soft background may be perfectly photographic and entirely intentional, while differing from the experience of standing there and looking around.

That accepted creative choice makes me cautious about using resemblance to ordinary sight as a test of photographic legitimacy. We rarely ask a portrait photographer to apologise because the background did not look quite like that to a person walking past. We understand the choice and what it contributes to the picture. The photograph describes the scene through an optical decision.

Exposure time makes the issue even clearer. A moving car’s lights can leave long trails in a photograph. Stars can form arcs through a long exposure, or through a sequence of exposures combined afterwards.⁠¹⁴ These arcs trace the stars’ apparent motion as the Earth rotates.⁠¹⁵ Those paths do not appear to our unaided eyes as complete luminous lines hanging in the scene. They become visible through recording over time.

A long exposure of flowing water similarly combines contributions from changing positions and reflections. The result may look silky, although the water was never a piece of silk. A short exposure can freeze a droplet into a shape that I could not examine during the original moment. Both images can arise from the same physical activity, represented over very different intervals.

Neither interval is automatically the truthful one. Human vision also has temporal limits, but it does not work like a camera shutter held open for several minutes. The photograph may record a fraction of a second, several seconds or much longer. We choose the interval that serves the picture or the measurement. ‘Real’ therefore needs to account for the time over which information was gathered.

Deep-sky photography extends this idea. Collecting faint light over time can reveal structures that are difficult or impossible to see unaided.⁠¹⁶ Some astronomical images also combine filtered measurements and assign colours to distinguish wavelengths.⁠¹⁷ Such images can communicate real observations without reproducing a view available to our eyes. Their meaning depends partly on understanding how the data were collected and displayed.

Movement and the photographic record

My own interest in intentional camera movement (ICM) makes these questions very tangible. When I move the camera during an exposure, a flower or tree can become a sweep of colour. The subject need not have moved in the way the photograph suggests. My movement changes where the subject’s projected light falls on the recording surface as the exposure continues.

The direction, speed and duration of my movement influence the marks that appear. My gesture becomes part of the exposure, giving me another way to interpret the subject. The result may resemble neither a conventional view of the scene nor what I saw while making it. Nevertheless, it is a record of an actual photographic encounter. The subject, the light and my movement all contributed. When I present it as ICM, I am inviting an artistic reading, and the description gives the viewer a fair understanding of how the image came about.

This brings me back to the three meanings of ‘real’. First, there is the physical subject and what happened to it. Second, there is the appearance experienced by an observer, which depends on the arriving light and the observer’s visual system. Third, there is the camera’s record, which depends on optics, exposure, the recording medium and subsequent processing. A single photograph can relate to all three, without matching them in every respect.

The old story of people touching different parts of an elephant comes to mind. Each person receives some information, but none has the whole animal from that encounter alone. I take this as a reminder of the limits of observation, rather than proof that there is no elephant. A photograph can give us something valuable and accurate while remaining partial.

Appearance and origin

Now imagine an AI system that produces a convincing image of a table. Suppose the design is new and no physical example has yet been made. It would be clearer to call this an imagined design than a reconstruction of an existing table. The image itself exists, and it may represent a perfectly buildable object. It simply does not document a physical table standing somewhere at that time. Next, suppose I use the design to make a table with a 3D printer. I then photograph it from the viewpoint shown in the generated image. For the purposes of this thought experiment, imagine that the results look identical. A viewer comparing the pictures alone might be unable to tell which came first or which was produced by a camera.

Both are real images, and both can accurately represent the table’s eventual design. Their histories, however, remain different. One image helped bring a table into existence; whereas the other recorded light from the table after it had been made. The later photograph does not retrospectively turn the earlier design image into a photograph.

If someone offered the first image as proof that the table had already been manufactured, the claim would be wrong, however accurately the image anticipated its appearance.

I would also separate existence from visibility. An object can exist without being seen by me, or by anyone at a particular moment. We can learn about things through measurements and indirect evidence as well as sight. The limits of our observation concern what we know about existence, but they do not automatically determine what exists. Otherwise, a table would have a rather unreliable life whenever we switched off the lights!

The thought experiment also helps distinguish synthesis from reconstruction. An AI system working from measurements of an actual scene might estimate missing information about something that existed. A system inventing a plausible table is doing something different. Both involve computation, but the relationship with observations matters. We should ask what data support a reconstruction, which parts are estimated, and how its accuracy has been checked.

The black-hole illustration Alan chose brings in another use of computation: scientific simulation. It is Jean-Pierre Luminet’s numerical simulation, published in 1979, of how a black hole surrounded by a thin accretion disc could appear.⁠¹⁸ Its scientific value comes from the physical model and assumptions behind it, and from how these can be tested against observations. An image which merely looks convincing has not demonstrated the same thing. Computation can produce a simulation, a reconstruction or an invented scene. We need to ask which kind of work it is doing.

What a capture establishes

In photography, a useful starting point is the encounter between light and a recording medium. In film, light creates a latent image in silver halide crystals, which chemical development makes visible.⁠¹⁹ A digital sensor converts incident light into electrical signals.⁠²⁰ Processing then turns those signals into a visible image.⁠⁷ The resulting image is connected to the conditions of exposure. It does not guarantee a complete or unaltered account of the subject.

There is a small catch here too. I could photograph an AI image displayed on a screen. The camera would record real light from the display, but that would provide no evidence that the depicted table had stood before the camera. Recording photons tells us about an encounter, but we still have to identify what was actually encountered. I would therefore ask what we mean when we say that a camera records ‘faithfully’. Faithful to what, and within which limits? Focus, exposure, sensitivity and processing all influence the result. Even an untouched image may omit something essential through its framing or timing. Photographic origin gives us an important kind of evidence, but the interpretation still needs care.

Intention and trust

For artistic work, I find the photographer’s intention especially helpful. Choosing a lens, aperture and exposure time is already a way of interpreting. Moving the camera deliberately adds a bodily gesture. The resulting picture may communicate my response to a place more successfully than a conventional sharp photograph. Its value can lie in that expression, while its photographic origin remains a separate, meaningful fact.

Being authentic to an intention does not make every claim true. I might sincerely want an image to communicate a memory of a place, yet still need to explain that I combined several moments. Equally, a documentary photographer’s care with evidence can be part of their artistic intention. The purpose of an image helps us interpret it, but intention and disclosure have to be considered together.

This has practical consequences for trust and competitions. The Nature definition of the International Federation of Photographic Art (FIAP) requires an honest presentation of the scene and restricts changes to image content, while allowing specified technical adjustments.⁠²¹ The RPS Digital Imaging Group’s Fun AI Competition invited AI-generated images in Photorealistic and Fantasy classes.⁠²² The same group’s 2026 DPI Open Annual rules accept montages and ICM, but forbid AI-generated source images and the introduction of significant visual elements absent from the entrant’s own photographs, while permitting AI editing only within stated limits.⁠²³ Permission depends on the competition’s published rules, not simply on whether a competition welcomes creative techniques. Entrants should check those rules and describe their process honestly. An appealing result cannot, by itself, establish eligibility.

So what does ‘real’ mean in photography? To answer that, we may first need to ask which relationship between an image and the world we have in mind. We can ask whether the subject existed, whether the image resembles someone’s visual experience, whether it records a particular encounter with light, and whether it honestly communicates its maker’s intention. A photograph can answer some of those questions very well without answering all of them.

The ultimate question

Perhaps, then, asking ‘What is real and what is not?’ brings us to another question: what do we mean by photography? This involves both science and philosophy. We can explain how an image is formed while still debating what it represents and what we can reasonably conclude from looking at it.

My working definition is that photography is the practice of producing images from recorded light or other radiation, with the result shaped by the conditions of capture and by how that record is processed and presented.

For me, its appeal is that this physical connection leaves room for personal interpretation. A photograph can reveal something beyond unaided vision, or express how a moment felt to the person making it. Neither possibility removes the need to be honest about how the image was made and what it claims to show.

An AI-generated image can also express a sincere artistic intention. However, looking photographic does not, by itself, establish that the depicted subject was recorded. That distinction remains useful even when the pictures are visually indistinguishable.

I therefore see photography as a meeting point between a physical record and the ways we choose to use and understand it. It can help us observe the world, explore it and express our experience of it. It does not have to reproduce exactly what our eyes saw to give us something worth seeing.

Photography can reveal aspects of reality, but being a photograph is not, by itself, a guarantee of truth.