Skip to main content

'I'm a consumer rights expert, here's what you should know about agentic AI'

We might soon have systems that autonomously plan, decide and act on your behalf
Hannah DownesSenior Writer

Hannah Downes covers consumer rights issues for Which?, helping to empower consumers and make their lives fairer

A woman with long black hair sits on the floor, using a laptop on a low table in a cozy, well-lit living room.

Online shopping habits are changing, with many people turning to AI chats, rather than Google search, to browse for products.

But while generative AI might still feel new, there are even more sophisticated agentic AI models being developed. These systems don't just answer prompts – they also perform pre-agreed tasks independently. 

A retailer's AI agent might, for example, repeat orders of milk or toiletries, placing these orders autonomously for you. They might also independently seek out deals and organise deliveries on your behalf.

For now, fully autonomous agentic AI is not readily available in the UK, largely due to regulations in the payment sector which require 'active consent' at the point of sale. But as a Consumer Rights expert, these are the things shoppers should be aware of when it arrives.

Be more money savvy

free newsletter

Get a firmer grip on your finances with the expert tips in our Money newsletter – it's free weekly.

Unsubscribe whenever you want. Your data will be processed in accordance with our Privacy notice.

Your rights on misleading or inaccurate information 

The AI tools we have access to now aren't always accurate. 

Last year,  our investigation into AI search tools such as ChatGPT and Copilot found they often made mistakes, misread information and even gave risky advice.

So what will this mean when retailers start using agentic AI? 

Under consumer law, retailers must give consumers the information they need to make informed purchase decisions. In March, the Competition and Markets Authority (CMA) confirmed these rules will apply to AI agents used by retailers, too.

It says all key information provided by AI agents – around prices, product information and your rights – must be accurate and in line with consumer law.

If a company uses an AI agent to provide a deal comparison service, for example, the AI agent must provide accurate results, as well as clearly disclose important information, such as how results are ranked and any links with suppliers.

How AI hallucinations are going to be managed

AI models can misinterpret data and 'hallucinate' results that are inaccurate.

The CMA suggests that businesses using AI should ensure there is a human actively checking that the AI agent is making correct decisions as part of the process.

If using AI for returns processes, for example, the CMA suggests that someone with appropriate experience should regularly review exchanges with consumers to make sure the AI agent is acting in line with the law and the retailer's policy.

It also advises companies to train their AI systems in consumer law.

Who is responsible if an AI agent breaches consumer law?

If a retailer's AI agent doesn't adhere to the rules outlined above – or if it makes it difficult for you to exercise your statutory consumer rights – it might be hard to tell who is responsible.

But the CMA says the retailer will be responsible, even if a third party designed the AI agent in question.

Retailers could be fined up to 10% of their worldwide turnover or face enforcement action if their AI agents break consumer protection law.

Crucially, the CMA states that businesses must act quickly if they spot that their AI agent is making mistakes or if there are problems.

The rules aren't clear on dodgy retailers and products

However, it's less clear who is responsible if an AI agent or platform, such as ChatGPT or Google Gemini, points you towards a fraudulent retailer or misrepresents a product.

Similarly, we don't know yet if payment protections such as Section 75 and chargeback would apply to purchases made after an AI recommendation, or within an AI platform.

With this in mind, Which? recommends always doing your due dilligence – and following the below tips – before making any purchases recommended by an AI platform or agent.

How to shop safely online

We always recommend that you follow these simple steps to ensure you're shopping safely – and getting the best price – online.

  1. Shop around for competitive prices and discount codes. AI agents might not necessarily be aware of websites listing additional discount codes for certain retailers.
  2. Pay with a credit card if possible. This means you'll have additional Section 75 protection (for items that cost between £100 and £30,000) if something goes wrong.
  3. Research the retailer. Check the company's social media pages and online reviews. If a retailer has had an influx of five-star reviews in a very short space of time, it's best to steer clear. Or if a retailer is offering a price that seems too good to be true, it's probably best to avoid it altogether. You can use our best shop reviews to find out what real customers think about a retailer before you buy.
  4. Research the product. Which? reviews a vast array of products from washing machines and fridges to hair curlers and LED face masks – and our research doesn't appear in AI chatbots. Our reviews are expert-led and impartial, so you can be confident you're investing in quality products.
  5. Check returns policies. Some retailers offer more generous returns policies than others, and many now charge return fees. It's worth checking if you'll be charged before placing an order.
  6. Check out as a guest if offered. Data breaches are on the rise. We always advise shoppers to check out as a guest – and to not save their card details with the retailer – if possible.
  7. Never specify a 'safe place' or a 'nominated neighbour' with couriers. Your rights are drastically reduced if you give these permissions and your parcel then goes missing.

Find out more: How to shop safely online