The Image That Was Never Shot: Licensing Synthetic Visuals for UK Campaigns in 2026
Artificial intelligence has moved from the edge of the studio workflow to the centre of campaign production. A UK marketer can now generate a plausible lifestyle scene, product backdrop or seasonal tableau in minutes, often before a traditional stock search has been properly briefed. The commercial attraction is obvious. The licensing question is less so. For UK designers, marketers and publishers, the test is not whether a synthetic image looks finished. It is whether its rights, provenance and accountability can survive procurement, legal review and the audit trail that follows a live campaign.
Access is not ownership
Traditional stock licensing starts with a relatively clear chain. A library identifies the asset, the contributor and the rights holder. The licence sets out permitted uses, restrictions, territories and durations. Where relevant, model and property releases support the use of people and places. Indemnities, where offered, sit within defined limits. Synthetic imagery breaks that chain into a different set of questions: which model produced the image, which version was used, what prompt and inputs were supplied, what human edits followed, and what the platform terms said on the day of generation.
Under UK copyright law, AI does not create a separate, simple ownership regime for every output. Human authorship remains central to most copyright analysis. The Copyright, Designs and Patents Act 1988 does contain a provision for computer-generated works, but it is not a general charter for owning model output, and its practical effect is limited and debated. Platform terms may grant broad commercial use, but a licence is not the same as ownership. Outputs can also be non-unique: similar prompts can produce similar results, and the same platform may generate comparable images for other users. A campaign team should therefore avoid the assumption that an AI image is automatically exclusive, automatically cleared, or automatically safe to indemnify.
Read the terms as part of the brief
AI platform terms vary materially between consumer, professional and enterprise tiers. Some services assign or license output rights to the user; others retain broader rights or impose restrictions on high-risk uses. Commercial use may be permitted, but the terms may exclude certain categories, require attribution, prohibit use in sensitive contexts, or limit indemnity to enterprise customers. The date matters. Terms change, models are retired, and a licence snapshot from last year may not describe the service used for this year’s campaign.
Indemnity deserves particular attention. Traditional stock licences often provide a defined indemnity against specified copyright or model-release claims, subject to caps and exclusions. AI platform indemnities are narrower and more conditional. They may cover generated output only, exclude prompts or reference images supplied by the user, exclude trademark or likeness issues, or require compliance with acceptable-use policies. If a client or publisher requires uncapped indemnity, or indemnity for all third-party claims, synthetic imagery may fail the procurement test before creative review begins.
Provenance and audit readiness
Stock imagery is generally audit-ready because the licence, asset ID, contributor and release status can be filed together. AI imagery needs a purpose-built record. At minimum, keep the platform and account, model name and version, generation date, prompt and negative prompt, seed or settings where available, reference images and their rights status, human edits, and a saved copy of the terms in force at the time. Record who approved the output and where it was used. If the image is later adapted for a second market or format, the audit record should show that the new use was considered.
This is not bureaucracy for its own sake. Campaigns are increasingly reviewed after publication by legal teams, procurement, platform auditors and clients’ own compliance functions. An image that cannot be traced back to a licence, a model version and an approval decision is a weak link in that review. A stock image with a clear licence may be less flexible than a synthetic alternative, but it is often easier to defend.
Disclosure, advertising and public trust
UK advertising rules require marketing communications to be obviously identifiable as advertising and not to mislead. Synthetic imagery is not automatically misleading, and there is no single blanket UK law requiring every AI-generated image to be labelled. Context is decisive. An abstract texture or clearly stylised background is different from a synthetic photograph presented as documentary evidence, a fake testimonial, or a scene implying that a real person, place or event was photographed. The Advertising Standards Authority and sector regulators may assess the overall impression, not the production method alone. Client policies, publisher guidelines and platform rules can add further disclosure requirements. If the campaign touches regulated products, public health, finance or news, legal review should come early.
Synthetic versus stock: a practical comparison
On risk, traditional stock is generally more mature: rights holders, releases and indemnities are established, even if limits apply. AI imagery carries greater uncertainty around provenance, similarity to protected works and the scope of platform assurances. On flexibility, AI wins: it can produce bespoke compositions, unusual angles and rapid variants. On audit readiness, stock usually wins unless the AI workflow has been documented deliberately. On cost, AI may reduce upfront production spend but can increase review time, legal checking and rework if the output cannot be cleared.
The sensible approach is rarely all or nothing. Use stock where the image makes a factual or documentary claim, where real people or identifiable places are central, or where a client needs a conventional licence and indemnity. Use synthetic imagery for abstract concepts, textures, mood, backgrounds and clearly stylised creative where the brief can tolerate its limitations. Hybrid workflows can work well, provided every element is logged.
A short clearance checklist
Before a synthetic asset goes live, confirm the following. Record the tool, model version, account and date. Save the prompt, settings and any reference inputs. Capture the platform terms in force and check commercial use, ownership, restrictions and indemnity. Identify the human authorship and edits involved. Run checks for trademarks, logos, recognisable characters, designs and locations. Confirm that no real person’s likeness is used without consent. Assess whether the context could mislead and whether disclosure is needed. Obtain client and legal approval where required. File the asset in the digital asset management system with an AI flag and a review date. Keep a cleared stock alternative available if indemnity or provenance requirements tighten.
Briefing with confidence
The brief should state whether synthetic imagery is permitted, where it may be used, what disclosure is expected and what evidence must be retained. The contract should allocate responsibility for prompts, reference inputs, platform compliance and final approval. Above all, teams should stop treating AI output as release-free stock. It is a different kind of asset with a different paper trail. Handled with that discipline, it can be a useful part of the UK campaign toolkit. Handled casually, it can create exactly the kind of rights gap that stock licensing was designed to close.
