Right of return for webshops in 2026: a 14-day cooling-off period does not mean every return is the same

Returns are treated as a logistical problem by many e-commerce companies.

Parcel comes back.

Warehouse checks it.

Refund.

Done.

But legally speaking, returning is much more complex.

There is an important difference between a commercial return policy that a webshop offers itself and the legal right of withdrawal of a consumer.

Anyone who mixes up those two may not give customers enough rights.

But also unnecessarily many.

And now that AI is increasingly assessing return requests automatically, a good legal setup is becoming even more important.

The basics: online sales usually have a 14-day cooling-off period

For online purchases, consumers generally have fourteen days to withdraw from the contract without giving reasons.

For products, that period in principle begins after the product is received. Different times apply for services.

This also means that as an entrepreneur you cannot require a consumer to first provide a detailed explanation of why they want to return.

You may ask.

But the answer must not be a condition for exercising the legal right.

“Sale cannot be returned” is not a general statutory exception

This is a classic example.

A webshop states for discounted products:

SALE — NO RETURN POSSIBLE.

Commercially understandable.

Legally not automatically valid.

The ACM explicitly confirms that the legal right of withdrawal can also apply to discounted items, showroom items and second-hand products.

An entrepreneur cannot therefore invent new legal exceptions themselves because a return is financially unfavourable.

But exceptions do exist

Not every online purchase can be withdrawn.

For example, clearly personalised products, rapidly perishable products and certain sealed hygiene products are subject to legal exceptions. For digital content, the right can lapse under specific conditions when the consumer explicitly consents to immediate delivery and the loss of the right of withdrawal.

But here too, there is an important nuance.

“Personalised” does not automatically mean that every configuration of a product is custom-made.

The ACM, for example, distinguishes between products that are actually made according to the specific wishes of the consumer and products that are only put together from standard options.

That difference is exactly why a simple AI rule like:

“custom = no return”

can be legally dangerous.

AI can automate returns — but also systematically reject them incorrectly

Suppose your webshop receives 1,000 return requests daily.

AI can analyse photos.

Categorise reasons.

Signal fraud patterns.

Prioritise returns.

Answer chat questions.

That can bring huge operational benefits.

But suppose the system automatically decides that the legal right of withdrawal does not apply based on words like “custom”, “opened” or “hygiene”.

Then one wrong legal rule can affect hundreds of customers.

That is the big difference between human errors and automated errors.

AI can make an error scalable.

That is why legal exceptions should not be put as separate prompts in a chatbot.

They must be translated into controlled decision rules, with human escalation where the situation is not clear.

An AI chatbot must not talk customers out of legal rights

This deserves special attention.

A customer asks via chat:

“Can I still send my order back?”

The chatbot answers:

“Unfortunately, your packaging has been opened and we do not accept opened items.”

That answer can be legally wrong.

For many products, a consumer is allowed to open the packaging and assess the product in a way that is comparable to what is possible in a shop. Only when further use leads to depreciation can compensation be considered under certain circumstances.

A chatbot that automatically communicates “opened = no return” therefore creates compliance problems itself.

And since 2 August 2026, the European AI transparency rules for directly interactive AI systems have also become relevant. Consumers must be able to recognise that they are interacting with AI in relevant situations.

That makes AI customer service both an operational and legal topic.

Since June 2026, something else has changed

For Dutch online sellers, there has been a particularly practical change since 25 June 2026.

The ACM indicates that entrepreneurs must offer an online function with which consumers can easily terminate an agreement during the cooling-off period: the so-called withdrawal button or cancellation function.

This function must be clearly discoverable and accessible during the cooling-off period. A company may only ask for data that is necessary to identify the purchase in question and must provide a confirmation after use.

That means that in 2026, a webshop does not only need a good return page.

The technical setup must also keep up.

The biggest mistake: creating as much friction as possible

Some companies try to reduce returns by deliberately making the process difficult.

Hiding the return form.

Requiring a login.

Asking five questions.

Talking to a chatbot first.

Then waiting for “approval”.

From a cost perspective, that thought is perhaps understandable.

From a compliance perspective, it can become problematic.

The direction of European consumer protection is precisely that online choices must not be made unnecessarily difficult by manipulative interface patterns. European regulators have already looked extensively at dark patterns such as hidden information and interfaces that guide consumers towards certain choices.

AI should therefore not be used to talk consumers out of their rights.

AI must be used to make a legally correct process more efficient.

AI can actually make returns much smarter

There is also a positive side.

When the legal basis is properly set up, AI can help distinguish between, for example, withdrawal, warranty, transport damage, wrongly delivered product and a commercial goodwill return.

That distinction is operationally valuable.

For example, a defective product does not simply fall under the same logic as someone who changes their mind within their cooling-off period.

AI can also signal patterns in return reasons.

If 18% of all returns for one SKU stem from “product smaller than expected”, the problem may not lie in your return process but in your product information.

With that, compliance suddenly becomes a source of business information.

Personalised pricing brings another AI question

AI is not only used for returns.

Webshops also use algorithms for pricing and personalisation.

When a price is personalised for a consumer based on automated decision-making and profiling, information obligations apply. EU consumers must be able to know before purchase when such a personalised price is applied.

That shows something bigger.

AI does not just touch one legal page.

It touches the entire customer journey.

From price.

To recommendation.

To checkout.

To customer service.

To return.

Not informing can become particularly expensive

Informing correctly about the right of withdrawal is important because the consequences of deficient information are not limited to a small text error.

According to the ACM, the cooling-off period can be extended by a maximum of one year in the event of insufficient information.

That makes this subject commercially relevant.

A mistake in a return page can therefore still have financial consequences months later.

Conclusion

A strong return process gives consumers their legal rights.

But no more than that because the webshop itself does not understand where the limits lie.

The correct structure distinguishes between statutory withdrawal, warranty, commercial goodwill and exception situations.

AI can speed up that process.

But the legal decision model must be right first.

Otherwise, you are not automating customer service.

You are automating errors.

Ecompliance

Do you want to check whether your return policy, checkout, withdrawal button, general terms and conditions and AI customer service legally align with each other?

Ecompliance can assess the entire process and translate it into a practical system that is both consumer-friendly and manageable from a business perspective.