AI product imagery means campaign and catalogue pictures made with generative AI from photographs of the real product, then art-directed, selected and retouched by people. For a premium brand it works well for settings, seasons and moods that would be costly to shoot, and badly wherever the picture itself is the evidence. This guide sets out where the line falls, the quality bar each frame should pass, and what UK advertising rules and the EU AI Act say about disclosure and rights.
This is general information, not legal advice. The rules are changing, so check the current position for your markets with an adviser.
What AI product imagery is
There are two quite different practices under the same name:
- AI-assisted campaign imagery. Real product photography, even simple phone shots, is used as the source. AI places the product in a new setting, light or styling, and people select, correct and finish the result.
- Fully generated imagery. The whole picture, product included, is generated from a description. For a brand selling a real product, this is rarely acceptable, because the product in the image is not the product customers receive.
This guide is about the first. The product stays real; the world around it is created.
When AI-assisted imagery works well
AI-assisted imagery earns its place where the setting matters more than proof:
- Settings that would be costly to reach. A rooftop at golden hour, a modern villa terrace, a winter scene in summer.
- Seasonal and market variations. The same product shown in settings that suit customers in London, Dubai or Munich.
- Campaign exploration. Testing several art directions before committing to a shoot.
- Social media volume. Varied, on-brand frames for a monthly plan without a new shoot each time.
- Catalogue lifestyle images. Context shots that sit beside true product photography rather than replace it.
When to shoot instead
Choose real photography wherever the image shows something customers rely on to decide:
- Results and efficacy. Skin before and after a treatment, the effect of a cosmetic, a clinical outcome.
- Exact finish and texture. Gemstone colour, the weave of a fabric, the grain of a leather, the sheen of a glaze.
- Fit and drape on a body. How a garment actually sits, moves and folds.
- Real places. Hotel rooms, villas, restaurants and clinics must look as they are, because customers book the place in the picture.
- Real people. Founders, staff, ambassadors and customers are photographed, with consent, never generated.
The UK Advertising Standards Authority has already ruled on a related case. In 2021 it found that social media ads for tanning products were misleading because beauty filters exaggerated the result, as its note on beauty filters in advertising explains. The same logic applies to any technique, AI included.
The quality bar for every frame
A premium AI-assisted image should pass the same inspection as a photograph. Check each frame at full size against these points:
- The product is true. Shape, colour, material, proportions and scale match a physical sample, not only the source photo.
- Logos and lettering are real. Marks, labels and packaging text come from your vector files and are composited by hand, never drawn by the model.
- Anatomy and edges hold. Hands, faces, jewellery clasps, hems and reflections are checked closely; distortions appear in exactly these places.
- Light is coherent. Shadows, reflections and colour temperature agree between the product and the setting.
- No borrowed identity. No other brand’s logo, product or signature design, and no recognisable real person.
- The set is consistent. Frames in one campaign share light, colour and styling, as if from one shoot.
- The file is fit for use. Resolution and crops suit every placement, from a website hero to a story.
If a frame fails one point, fix it or discard it. Volume is never the goal.
Disclosure and the rules
UK: advertising codes apply however the image is made
In the UK, the CAP Code is media-neutral. In its June 2026 note, AI and deepfakes: four things advertisers need to know, the ASA confirmed that the rules apply regardless of how content was created, and that advertisers remain responsible for AI outputs.
On disclosure, the ASA’s May 2025 article on disclosure of AI in advertising states there is no blanket legal requirement in the UK to disclose AI use in ads. The test is whether the ad would mislead. A label can clarify that content is not meant literally, but adding “made with AI” does not cure an image that misrepresents a product, such as an AI picture exaggerating what a cosmetic achieves.
The ASA also notes that it does not regulate intellectual property or personality rights, so those questions need separate legal advice.
EU: Article 50 of the AI Act
The EU AI Act (Regulation (EU) 2024/1689) adds transparency duties that apply from 2 August 2026. Under Article 50(4), anyone using AI to generate or manipulate image, audio or video “deep fakes” must disclose that the content is artificial. The Act defines a deep fake as content resembling existing people, objects, places, entities or events that would falsely appear authentic or truthful.
A photorealistic image of a real product in a convincing real-looking place may fall within that definition. Where content is part of an evidently artistic, creative or fictional work, the duty is reduced to disclosing that such content exists, in a way that does not spoil the work. Separately, Article 50(2) requires the providers of generative tools to mark their outputs in a machine-readable way.
A practical labelling policy
Many brands settle on one consistent approach rather than deciding case by case. A sensible policy:
- Credit AI-assisted campaign imagery in a quiet, consistent line, for example in the caption or the site footer.
- Label photorealistic AI images of people clearly wherever they run, and never present them as real customers, staff or ambassadors.
- Never rely on a label to excuse an image that misrepresents the product.
- Follow each platform’s own AI labelling rules and keep them under review.
Rights in AI images
Ownership of AI-generated images is unsettled, so build the protection into your process and contracts.
- Copyright. UK law contains a provision for computer-generated works, but the government’s report on copyright and AI, published on 18 March 2026, proposes removing protection for works generated wholly by AI while keeping it for works created with AI assistance. Human direction, selection and retouching therefore matter, so keep a record of them.
- Inputs. Use only material you own or are licensed to use: your product photography, your logos, licensed references.
- Tool terms. Use tools whose terms allow commercial use, and check how they treat your uploads, including whether inputs may be used for training.
- Likeness and marks. Do not imitate real people, their voices, or other brands’ logos, products or signature designs.
- Contracts. Agree who owns the final images, where they may be used, and what the studio or agency warrants about tools and inputs.
A workflow that keeps quality high
- Brief. An art director sets the setting, light, styling and mood, in line with the brand.
- Source. Photograph the real product cleanly from the angles the campaign needs.
- Generate. Produce options; a person selects, rejects and refines them.
- Retouch. Correct the product against a physical sample and composite logos and lettering from real files.
- Check. Inspect every frame at full size against the quality bar.
- Label and record. Apply your labelling policy and keep a note of tools, inputs and the human steps.
Imagery is one part of a wider plan; our guide to social media strategy for premium brands covers how campaign images fit a monthly feed.
AI product imagery checklist
- The image is not evidence of a result, finish, fit, place or person; otherwise, shoot.
- The source is real photography of the real product.
- Shape, colour, material and scale match a physical sample.
- Logos and lettering are composited from vector files.
- Hands, faces, edges and reflections pass a full-size check.
- No third-party marks, designs or recognisable people.
- The labelling policy is applied, including platform and EU rules where relevant.
- Tool terms allow commercial use; inputs are owned or licensed.
- Ownership and usage are agreed in writing.
- A record of tools, inputs and human steps is kept.
Where to start
Pick one product and one campaign idea that would be hard to shoot, and test an AI-assisted set against the checklist before scaling up. Compare it with your current photography on the website and feed, not on a blank slide.
Workly Up’s AI creative work follows these rules, and our AI principles set out how we label, check and handle rights on every project.
