AI-Generated Ads Are About to Change Online Shopping — But Should Every Image Be Labeled?

AI-Generated Ads Are About to Change Online Shopping — But Should Every Image Be Labeled?

Artificial intelligence is changing online advertising faster than many consumers realise.

A growing number of retailers are using AI to create product images, design marketing campaigns, generate virtual models and produce advertising content at a fraction of the traditional cost. In some cases, AI can create an entire visual campaign without a photography studio, a physical location, a large creative team or even a real human model.

For businesses, the attraction is obvious. AI can make marketing faster, cheaper and easier to personalise. For consumers, the results can be harder to see.

A fashion image may look like a normal photo. A living-room advertisement may look like it was shot in a real home. A model displaying a jacket may appear human even when the person does not exist. A product image may be edited so heavily that it no longer reflects what buyers will receive.

That is why the European Union is preparing new transparency requirements for AI-generated content.

The goal is simple: consumers should be able to understand when artificial intelligence has played an important role in creating or changing what they see.

But the practical question is much more complicated.

Should every AI-generated advertising image be labeled?

Retail groups argue that a broad labeling rule could create confusion and lead to warning fatigue. They say consumers may begin seeing AI labels everywhere, including on harmless edits, generic product backgrounds or marketing images that are not meant to deceive anyone.

Consumer advocates take the opposite view. They argue that if an image is generated or significantly altered by AI, people deserve to know. Without disclosure, shoppers may assume that they are looking at a real product photo, a real person or a real place.

The debate is becoming more urgent because AI advertising is moving from experimentation to normal business practice.

Retailers already use AI to improve lighting, remove backgrounds, create variations of products, translate campaigns into different markets and generate images for social media. These are relatively simple uses. But the technology is also being used to create complete scenes, fictional customers and artificial models.

That changes the relationship between advertising and reality.

Traditional advertising has always involved editing. Photos are retouched. Products are styled. Lighting is adjusted. Commercials use actors, sets and visual effects. Consumers generally understand that advertisements are designed to present products in the best possible way.

AI makes that process more powerful because it can create visual material that never existed in the first place.

A company may no longer need to photograph a real sofa in a real room. It can generate a perfectly designed living space around a digital version of the product. A fashion retailer may no longer need to organise photo shoots with different models, photographers and locations. It can generate dozens of campaign images quickly and adapt them for different audiences.

This can reduce costs dramatically.

For smaller businesses, that may be a major opportunity. A small online store could create high-quality marketing material without spending thousands of euros on a professional campaign. A local brand could test different designs, seasonal looks or social-media concepts without needing a large agency.

AI could make marketing more accessible.

But lower costs can also lead to more content, more manipulation and less clarity about what is real.

That is where the EU’s transparency rules come in.

The European Union’s AI Act is designed to create rules for different types of artificial intelligence, especially where AI may create risks for safety, privacy, consumer protection or public trust. Under upcoming transparency requirements, certain AI-generated or AI-altered content may need to be clearly identified, particularly when it could be mistaken for authentic material.

The intention is not to ban AI-generated advertising. It is to make sure consumers are not misled.

However, retailers argue that the wording and scope of the rules could create uncertainty.

A retail association representing companies such as Amazon, H&M, Inditex and Ikea has argued that AI-generated ads not intended to deceive consumers should not automatically be treated like deepfakes. The group’s position is that a generated image of a sofa in a living room, for example, should not necessarily require the same type of warning as a manipulated video designed to falsely show a public figure saying something they never said.

This distinction matters.

Deepfakes are usually understood as realistic synthetic media that can deceive people about a real person, event or statement. A fake video of a politician, a celebrity or a journalist can create serious harm because it can spread false information while looking authentic.

An AI-generated product scene may be different. It may be artificial, but it may not necessarily be trying to trick the viewer about a real-world event.

Still, consumers may see the issue differently.

If a shopper sees a model wearing a dress, they may assume the image is a real photograph. If the model is entirely generated by AI, should that fact be disclosed?

If a buyer sees furniture in a beautiful apartment, they may assume the product has been photographed in a real setting. If the scene is artificial, should the company say so?

If an image shows a cosmetic product producing a certain effect on skin, should companies disclose whether the visual result was enhanced by AI?

These questions show why the debate is not only about technology. It is about trust.

Advertising works because consumers believe that a product can deliver something close to what the campaign promises. If AI makes images look more realistic, more polished and more persuasive than real photography, companies may gain a powerful new tool. But they may also create higher expectations that products cannot meet.

This is especially sensitive in fashion, beauty, home design, food and travel.

A digitally generated fashion model may have perfect skin, perfect proportions and clothing that appears to fit perfectly. A real customer may receive the same product and find that the fit, colour or material looks different.

A digitally generated holiday image may show an idealised beach, hotel room or restaurant that does not exist in reality. A buyer may then make decisions based on an image that is technically impressive but misleading in practice.

The same applies to product photography.

AI can sharpen details, improve colours, remove imperfections and create visual effects that make an item appear more expensive or more attractive than it is. Some level of editing has always existed, but AI makes the process easier to scale across thousands of products.

That creates a challenge for regulators.

Rules must be clear enough to protect consumers, but flexible enough to avoid treating every small edit as a major deception risk.

If companies are forced to label every image touched by AI, the result could become meaningless. Consumers may see labels on product photos, social posts, websites and online ads so often that they stop paying attention.

But if rules are too weak, companies may use AI to create highly persuasive content without giving people enough context.

The question is where to draw the line.

One possible approach is to focus on material changes.

A company may not need to label a basic AI-generated background if the product itself is shown accurately. But a label could be required when AI changes a product’s appearance, creates a fake human model, simulates a real-world event or alters results in a way that could influence a customer’s decision.

This would focus disclosure on situations where AI meaningfully affects consumer understanding.

Another approach is to use simple, standardised labels rather than large warnings.

For example, online retailers could use a small icon or short text such as “AI-generated visual” or “AI-enhanced image.” The label could be visible without making the page unreadable or suggesting that every AI-made image is dangerous.

The success of this system would depend on consistency.

If every platform uses different language, consumers may not understand what the labels mean. One company may write “synthetic media,” another may write “AI assisted,” and another may simply use a small symbol. That could create even more confusion.

A clear common standard would make it easier for consumers to understand when a visual is real, edited or generated.

The wider issue is that AI is changing the economics of online content.

Retailers can now create more campaigns faster. Brands can test ads for different audiences. Marketing teams can produce images in many languages and styles. Small companies can compete visually with larger brands.

This could make e-commerce more creative and competitive.

But it also means the internet may become filled with content that looks polished without being connected to real people, places or products.

Consumers may eventually become more sceptical. They may ask whether an image is real, whether a review was written by a person, whether a model exists or whether a product scene has been generated entirely by software.

That is not necessarily a bad thing. More critical thinking can help people make better decisions online.

But trust is valuable. If consumers begin assuming that every image is manipulated, brands may find it harder to build credibility.

For retailers, the long-term question is not only whether AI can save money. It is whether AI can be used without damaging trust.

The most successful companies may be those that use AI openly and responsibly. They can use the technology to create better experiences while still being honest about where visual content comes from.

This may become a competitive advantage.

A brand that clearly labels AI-generated content could be seen as more transparent. A retailer that uses AI to show products in different styles or settings could still make clear what the buyer is actually receiving. A company that uses virtual models could explain why: to create more inclusive campaigns, reduce production costs or show more product variations.

Transparency does not have to make advertising less effective.

In some cases, it may make advertising more credible.

The EU’s new rules will test whether companies, regulators and consumers can find a workable balance. The goal should not be to stop innovation. AI-generated advertising is likely to become normal.

The goal should be to ensure that people are not left guessing what is real.

Online shopping already depends on trust. Customers cannot touch a product, try it on or inspect it in person before buying. They rely on images, descriptions, reviews and brand reputation.

If those visual signals become increasingly artificial, clear disclosure may become more important, not less.

The future of online advertising may not be about whether AI is used.

It will be about whether people can still trust what they see when it is.