AI visual compliance
AI product images and the law: a practical guide
Using AI is not what creates legal risk. Regulators ask an older, simpler question: does the page mislead a shopper about what they will actually receive? Here is how that applies to product imagery — and how VizIQ is built around it.
Last reviewed
Is it legal to use AI-generated product images?
Yes. Across the US, the EU and Israel, existing consumer-protection and advertising law already covers images, edited media and implied claims. There is no AI exception to truth-in-advertising rules, and equally no rule against generating imagery. AI raises the stakes only because it produces persuasive visuals at volume.
What does not work is a disclaimer bolted onto a misleading picture. In April 2026 the UK's Advertising Standards Authority considered an ad carrying the label "AI-generated … for illustrative and branding purposes only", acknowledged the label, and then wrote "Nevertheless" before upholding the complaint on the overall impression the ad created. Accuracy is the control that matters. Disclosure supports it; it does not replace it.
A rule of thumb
Not all AI imagery carries the same risk
A practical way to classify what you publish, aligned with how regulators actually assess advertising.
AI-assisted edits
Background cleanup, relighting, recropping, shadows, upscaling. The product itself is unchanged, so the shopper's expectation is unchanged. Generally no special disclosure is expected.
Real product, generated scene
Your validated product shown in a synthetic room or styling. The product is true; the context is constructed — and context shapes expectations. Disclose that the scene is generated.
Synthetic humans, or an altered product
Generated colour, texture, packaging or accessories — or a photorealistic AI human. Disclose prominently, keep a real product image alongside, or don't publish it.
What changed in 2026
The direction is toward provenance, not away from it
EU AI Act Article 50 applies from 2 August 2026
Transparency obligations are now live. The European Commission's guidelines of 20 July 2026 treat a realistic AI-generated person as in scope even where that person is entirely invented — their own worked example is a synthetic influencer presenting a real product. Disclosure that can be "easily overlooked or missed" does not count. Note the duties split: the tool provider carries machine-readable marking, and the brand publishing the ad carries visible disclosure. They are not interchangeable.
New York requires synthetic performer disclosure
General Business Law §396-b took effect on 9 June 2026. Advertising that features a synthetic performer — an AI-generated human who is not any identifiable real person — must conspicuously disclose it, where the advertiser has actual knowledge. Enforced by the Attorney General.
Amazon now mandates a metadata tag
Since 23 July 2026, listing images, video and A+ content containing photorealistic AI-generated people must carry a metadata keyword, worldwide, with a shopper-facing indicator. Images of real people — even AI-edited ones — are outside the rule. Google Merchant Center separately requires generative-AI images to carry IPTC origin metadata and asks that you not strip embedded tags.
How VizIQ is built
Every image starts from your real product
VizIQ is a closed-world system: generation begins from your own product photo and its metadata, never from a text prompt alone. That single design choice means there is always a validated reference behind the output — the thing a prompt-only tool cannot offer. On top of it sit four controls.
A human signs off before anything ships
ReviewSend a batch to any stakeholder — including people without a VizIQ seat — and they approve or reject each shot from a phone. Nothing publishes on its own.
Every decision is logged against the SKU
AuditApprovals and rejections are recorded server-side with the SKU, the reviewer and the timestamp. When someone asks who checked that a shot was still accurate, there is an answer.
Your source image stays the reference
OrganizerThe original photo is retained and can anchor the set as the controlling product image, so the truth layer is never lost behind the generated ones.
Copy changes are drafts until you accept them
EnhanceAI never writes to your product text directly. Every field lands as a draft with its origin recorded — imported, AI-inferred or human-edited — and the original merchant value is kept.
Human models
Our models are our own
The curated human models in VizIQ's fashion library are VizIQ's own creation — not scraped, not licensed from a third-party dataset, and not derived from photographs of identifiable real people. There is no publicity-rights or likeness-consent chain for you to administer, and no real person whose permission could later be contested.
They are, of course, AI-generated people — so where a channel or jurisdiction requires you to disclose a synthetic performer, that obligation still applies to what you publish. Owning the model answers who holds the rights. It does not remove the disclosure step, and we would rather say so plainly.
Common questions
Are AI-generated product images legal? +
Yes. No jurisdiction bans AI-generated product imagery. What is regulated is whether the overall impression misleads a shopper about the product they will receive, and — increasingly — whether AI-generated humans are disclosed.
Is "for illustration only" enough? +
No. Regulators reject generic disclaimers where the image itself misleads. The UK's ASA has explicitly acknowledged an AI-generated disclaimer and upheld a complaint anyway, on the overall impression of the ad.
Whose likeness is in the models? +
Nobody's. VizIQ's curated fashion models are our own creation and our own IP — not derived from photographs of identifiable real people. You can also bring your own model imagery, in which case you are responsible for holding the necessary rights and releases.
Is VizIQ output marketplace-compliant? +
VizIQ generates clean product imagery without the on-image text, logos or watermarks that major marketplaces restrict in listing photos. Each marketplace sets its own policy and they differ — for example, some restrict only the primary image — so check the policy for the channel you publish to.
Does VizIQ make my brand compliant? +
No tool can. Under the EU AI Act the obligations split: the provider of the generation tool and the brand deploying the imagery each carry their own duty, and they cannot be substituted for one another. What VizIQ gives you is the accuracy, review, audit trail and reference imagery you need in order to meet yours.
Not legal advice
This page is a practical summary for e-commerce teams, not legal advice, and the law here is moving quickly. Several questions — including whether New York's synthetic performer rule reaches static product photography — are genuinely unsettled and competent lawyers disagree. Please have counsel review your own disclosures before publishing.
Bring this to your legal team
If a compliance review is holding up your visual content, we are happy to walk your counsel through how generation, review and audit actually work.