When AI Photography Is Not the Right Answer

The situations where a camera is cheaper, safer or simply required, and how to recognise them before commissioning generated imagery.

Generated imagery has become good enough that the interesting question is no longer whether it works but where it should not be used. A studio that offers it for everything is not advising, and clients who adopt it uniformly encounter the same set of problems in the same order. The boundaries are identifiable and worth stating plainly.

The first and firmest boundary is documentary evidence. An image asserting that something exists, a facility, a team, a product in a customer's hands, an event that happened, has to be a photograph. Generation produces the plausible rather than the actual, and the gap between those is exactly what the audience is relying on. This is not a quality threshold that better models will cross.

The second is identifiable real people. Staff, founders, customers and spokespeople must be themselves, and any workflow that starts from photographs of a specific individual requires their documented, revocable consent. The U.S. Copyright Office (2024) addressed the digital replicas question directly in its report on that subject, and the exposure is legal as well as reputational.

The third is anything the buyer will physically receive. Product colour, texture, finish, proportions and included accessories must match what arrives, because the customer can check. Kong and Lou (2026) found that visual appeal and visual congruence play distinct roles alongside social proof in how advertising is processed, and in commerce the congruence failure appears directly as a return, a refund and a review.

The fourth is regulated claims. Medical, financial, safety and food categories require that what is shown corresponds to what is approved or substantiated, and a reviewer will ask where an image came from. Generated material can be illustrative in these settings and cannot depict outcomes, procedures or results.

The fifth is a commercial rather than an ethical boundary and is frequently overlooked: exclusivity and protection. The U.S. Copyright Office (2025a) concluded that copyright protects human authored expression in works made with AI tools, that outputs lacking meaningful human creative input do not qualify, and that prompt selection alone does not by itself produce a copyrightable work. A brand that needs to prevent a competitor reusing its imagery has a stronger position with substantially authored work.

There is also a straightforward cost boundary that gets missed because generation feels cheap. If a product is in the office, a photographer for half a day produces accurate images of it faster and at lower cost than building a convincing generated version with composited artwork. Generation earns its place where photography is expensive or impossible, not where it is simply available.

The audience perception boundary is softer and real. Kirk and Givi (2025) found that perceptions of AI authorship shape consumer responses to marketing communications and can produce negative reactions in some conditions, and Farooq and de Vreese (2026) documented how awareness of AI generation affects authenticity judgements. In categories where the proposition is craft, provenance or human care, the method itself can undercut the message even when the image is good.

Against all of this sits the substantial territory where generation is clearly correct: environments that would require travel or permits, products that do not exist yet, concepts and future states, abstract and atmospheric material, visual exploration before commitment, and campaign worlds that are understood as constructed. In these the alternative is not a better photograph, it is no image at all or a much more expensive one.

The practical test to apply before commissioning is a single question asked of each image: does this assert something about the real world that a viewer could check. If yes, photograph it. If no, it is a candidate for generation. Applying that test image by image, rather than adopting a blanket policy in either direction, is what produces a library that is both affordable and defensible.

References

U.S. Copyright Office. (2024). Copyright and artificial intelligence, Part 1: Digital replicas. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf

Kong, J., & Lou, C. (2026). Beyond persuasion knowledge: Examining the roles of visual appeal, visual congruence, and social proof in influencer advertising. Journal of Retailing and Consumer Services, 88, Article 104502. https://doi.org/10.1016/j.jretconser.2025.104502

U.S. Copyright Office. (2025a). Copyright and artificial intelligence, Part 2: Copyrightability. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf

Kirk, C. P., & Givi, J. (2025). The AI-authorship effect: Understanding authenticity, moral disgust, and consumer responses to AI-generated marketing communications. Journal of Business Research, 186, Article 114984. https://doi.org/10.1016/j.jbusres.2024.114984

Farooq, A., & de Vreese, C. (2026). Deciphering authenticity in the age of AI: How AI-generated disinformation images and AI detection tools influence judgements of authenticity. AI & Society, 41(1), 493–504. https://doi.org/10.1007/s00146-025-02416-5