Photo editing ethics: what you can and cannot change

Every photograph is edited in some way. The question is whether the edit changes what the picture says happened. For organisations that publish field photographs, that question goes to the heart of trust. Bond's ethical guidelines for NGOs say audiences need to be able to rely on images as credible evidence that something happened.¹

Bond accepts that designers and editors will make some post-production changes, such as cropping for layout. It states plainly that it is not acceptable to make any changes to a photograph or film that change its context or meaning.¹ That single test, does this edit change what the viewer will believe?, is the most useful rule a communications team can adopt.

Documentary photography has a well-tested line, and the World Press Photo contest sets it out clearly. Its guidance on what counts as manipulation allows colour adjustments and conversion to black and white when they do not alter content, cropping that removes extraneous details, and the removal of sensor dust or scratches on scanned negatives.²

Its entry rules also allow smart or AI-powered tools such as noise reduction, automatic adjustments to levels, colour and contrast, and object selection for local adjustments, but only as far as they do not alter, add or remove information.³ In practice, the following are normally safe for documentary and NGO photographs:

• Correcting exposure, white balance and contrast.
• Converting to black and white.
• Cropping to remove distracting edges, without hiding context.
• Reducing noise and sharpening moderately.
• Removing sensor dust spots.

World Press Photo's entry rules say a photograph must not be altered by adding, rearranging, reversing, distorting or removing people or objects within the frame.³ Its manipulation guidance gives concrete examples: erasing marks on a person's body, small objects, reflected light spots or shadows all count as manipulation, as do cloning in highlights, painting in details and photo montage.²

Tone and colour have limits too. The same guidance rules out colour changes so large that the processed colours diverge from the original, and changes to density, contrast, colour or saturation that significantly alter content, especially where they obscure or remove information.² Darkening a background until a crowd disappears, or turning a grey sky dramatic to suggest a crisis, crosses that line.

Ethics in editing is not only about pixels. It is also about how people are shown. Bond's guidelines ask organisations not to crop, edit, colour or do anything else that enhances the perception of poverty or abandonment.¹ A tight crop that removes a parent standing beside a child, or a desaturated grade that makes a busy market look bleak, can mislead as surely as removing an object.

World Press Photo's code of ethics asks that the content and edit of photographs give an accurate and comprehensive representation, and that photographers avoid reinforcing stereotypes.⁴ In our own practice, we compare the final edit with the original file and ask whether the people photographed would recognise their situation in it.

Some photographs are directed: a portrait, a reconstruction of an event, a scene set up for a campaign. That is not wrong in itself, but it must be honest. World Press Photo's code says photographers must not intend to mislead by recreating or staging events, and that any direct influence over the scene, such as re-enactments or posed portraits, must be included in the caption.⁴ It also requires captions and other text to be accurate.⁴

Bond takes the same view for NGOs: when images are set up for creative or fictional purposes, the audience should be told, and they should never be presented as reality.¹ Note that phone cameras process images heavily. World Press Photo accepts smartphone photographs only when taken in standard mode, and does not accept HDR, portrait mode, creative lighting effects or panorama mode.³

Generative AI has made photo manipulation faster and harder to see. World Press Photo does not allow synthetic or AI-generated images in its photography categories, bans generative fill, and treats AI enlarging tools as a breach of its rules.² Bond does not recommend AI-generated images except for budget reasons or to protect people at risk in sensitive contexts, and says AI imagery is not a way to avoid consent or solve problems of representation.¹

When you do use AI images, Bond says it is critical to make it clear and obvious, for example with a caption stating that the image is AI-generated.¹ Embedded metadata helps too. The IPTC Digital Source Type vocabulary distinguishes a camera capture, an algorithmically enhanced photo, an image edited with generative AI through inpainting or outpainting, and media created entirely by a trained AI model.⁵ Keep that metadata, but do not rely on it alone: Google warns that credit and source metadata can be modified by editing tools.⁶

The best protection against accusations of manipulation is evidence. World Press Photo asks finalists for the original camera files, such as RAW files, full-format JPEGs, unedited smartphone images or unprocessed scans, and two independent digital analysts compare them with the entries to check whether content has been altered.⁷

Organisations can adopt a lighter version of the same discipline. Archive the original file next to every published version, record who edited it and what was changed, and write a short internal policy that lists acceptable and unacceptable edits. When a journalist, donor or community member asks whether a photograph is real, you can then show them.

• Colour, exposure and contrast corrected without changing content
• Crops checked so they do not hide context or exaggerate need
• No people or objects added, removed, moved or distorted
• Staged or posed scenes disclosed in the caption
• Captions checked for accuracy: who, what, where and when
• AI-generated images visibly labelled and metadata kept
• Original files archived next to published versions
• A written editing policy shared with staff and suppliers