Shoppers embrace AI assistants but ditch poor advice

Most consumers in the UK and US are open to using AI assistants for online shopping, but many quickly walk away when recommendations miss the mark, according to new research commissioned by Nosto.

A survey of 2,000 consumers found that 66% have tried, or would be open to trying, shopping online with an AI assistant such as ChatGPT or a retailer's own tool. However, 69% of those who received irrelevant product suggestions stopped using the assistant and searched elsewhere.

The results suggest retailers face a high bar for relevance. Shoppers also want clarity on how their information is used and do not want automated systems taking control of buying decisions.

Expectations rise

Across the sample, 72% said they expect AI shopping assistants to help with online shopping. Consumers cited use cases ranging from deal hunting and gift ideas to fashion guidance.

Deal and price-drop alerts ranked highest among desired features (59%). Personalised recommendations followed (51%), then gift inspiration (44%). Real-time Q&A, such as "What goes with this?", came in at 41%. Bundles that save money or complete a look ranked at 40%.

While large consumer platforms have taken early steps into AI-led shopping, many shoppers are also interested in retailer-owned assistants. In the survey, 48% said they would be open to using an AI shopping assistant from their favourite retailer if it launched this year. Openness rose to 62% among 25- to 34-year-olds.

Trust and control

Concerns about data use and user agency emerged as barriers. The top reasons consumers gave for stopping use were worries about how their data is handled (24%) and the feeling that AI is making decisions without their input (21%).

Nosto CEO Jim Löfgren said adoption depends on execution, including safeguards and responsible data practices.

"The high level of AI acceptance sets the stage for rapid adoption for the right use cases, and for retailers it's all about execution now. That means delivering guardrailing and training the agent to ensure relevance from the first interaction, and ensuring that the data is protected and used responsibly. The brands that get this right will win!"

Early adopter behaviour

The study found that 34% of consumers have already tried shopping with a conversational AI assistant, either on a brand's website or through a general conversational interface. Among 25- to 34-year-olds, the figure rose to 59%.

Experience appears to change attitudes: among those who have tried AI shopping assistants, 77% said they would trust a brand more if it offered one.

Early adopters also expressed clear preferences for how AI should appear on retail sites. When asked how they would prefer to access an assistant for product discovery, 46% chose a website's search bar, while 41% preferred a dedicated on-site chatbot.

The result suggests retailers may need to treat AI as part of the search experience, rather than an add-on conversation window. It also underscores the risk of a poor first interaction: after receiving irrelevant suggestions, 69% of early adopters said they gave up and searched elsewhere. Only 28% said they rephrased their query and continued using the assistant until it found something relevant.

Löfgren also pointed to the need for richer retail data behind the interface.

"This makes personalisation more important than ever," he said. "As LLM-powered AI assistants become a new interface for shopping, their value will increasingly depend on how well they are trained on rich, intent-driven and contextual commerce data, rather than on natural language capabilities alone. This is what enables them to deliver truly personalised responses and product recommendations that turn conversations into conversions."

Discovery over checkout

The research suggests consumers see AI's biggest impact earlier in the shopping journey. Among early adopters, 33% said AI can have the greatest impact by improving decision-making, such as comparing or choosing products. Another 25% pointed to product discovery-finding the right items.

Checkout ranked lower: 20% of early adopters said improving payment speed and ease would be where AI makes the biggest difference. Post-purchase service, including returns and support, came in at 17%.

As shoppers become comfortable with AI, they are more willing to use it for more complex tasks, the survey suggests. Among early adopters, 81% said they would be open to using AI to build a full shopping cart for occasions such as holidays, back-to-school, or gifting. Another 88% said they would find it helpful if AI recommended bundles or complementary items.

Retail spending gap

Nosto also surveyed 39 eCommerce retailers about investment in AI agent tools. It found that 36% currently allocate additional, dedicated budget to these tools. Among those investing, three-quarters said their current investment is limited.

Löfgren said there is a mismatch between what consumers want and what many retailers are funding.

"The readiness gap between consumer demand and retailer investment in AI is not massive, but it's real. While shoppers stand ready to embrace AI, many retailers have been slow to respond. Those who move first and focus on transparency, relevance, and ensuring shoppers feel in control of their buying decisions, are likely to gain a clear competitive advantage."