The Shopping Journey Is Splitting: 20% of UK Consumers Now Start Product Discovery With AI

The World Retail Congress and Vercel findings also reveal an important tension between influence and trust. Reporting on the research suggests that 87% of UK shoppers find AI recommendations useful, while 55% say they have purchased something following an AI recommendation. Those figures indicate that AI is already playing a meaningful role in purchasing behaviour, even if the consumer does not complete the transaction directly inside an AI interface.

At the same time, shoppers are not simply handing every buying decision to an algorithm. Only 10% of respondents reportedly preferred spending time with an AI shopping assistant over a human performing the same role, while around a third were concerned that AI might not surface the best products, particularly if commercial incentives influenced what appeared. The emerging behaviour therefore looks more like assisted decision-making than complete delegation. AI can reduce thousands of possible products to a shortlist, but customers may still verify the recommendation through reviews, retailer websites, independent coverage or their own research before buying.

That distinction matters because it shows where AI currently sits in the customer journey. It does not necessarily need to complete the sale to influence revenue. If an AI assistant determines which three or four brands a customer investigates next, appearing in that initial shortlist can be commercially significant even if the final conversion happens elsewhere.

AI visibility can sometimes sound like a branding metric: does ChatGPT mention the company, does Gemini understand what it does and does an AI answer cite its website? Retail makes the commercial significance easier to see because recommendation systems can directly influence which products a customer considers. Imagine someone asks: “What are the best trail running shoes for a beginner who mostly runs on wet UK footpaths and wants to spend less than £120?” The brands surfaced in that answer have entered the buying process. Brands that do not appear may never be evaluated.

That changes how marketers should think about visibility. A retailer could perform strongly in Google for traditional product keywords and still be largely absent from conversational product recommendations. Conversely, a product may begin appearing frequently within AI answers before significant referral traffic is visible in analytics. The measurement framework therefore needs to expand beyond rankings and clicks to include questions such as whether a brand appears for important use cases, which competitors are recommended instead and whether product information is represented accurately.

For retailers, one of the clearest developments is the growing importance of structured, current product information. OpenAI provides infrastructure allowing merchants to share product feeds so ChatGPT can more accurately index and display products, including information such as pricing and availability. That points towards a wider direction of travel: AI shopping systems need reliable information that can be confidently interpreted and compared.

A vague product page may contain enough information for a human customer who is prepared to browse several tabs, inspect images and make assumptions. An AI system evaluating dozens of products needs clearer evidence. Product names, specifications, sizes, colours, prices, availability, shipping details, use cases and other relevant attributes increasingly form part of a retailer's visibility infrastructure. The quality of the information underneath a campaign may therefore become just as important as the campaign itself.

Content strategy also needs to reflect the way people actually ask AI systems for help. Ecommerce websites are traditionally organised around categories and product pages, but conversational discovery often begins with a problem rather than a product name. A customer might ask for a waterproof jacket suitable for commuting and weekend walks, a laptop for video editing under a certain budget or skincare suitable for a particular set of requirements. Product pages and supporting editorial content need to answer those real buying questions clearly, explaining who products are for, what differentiates them, which trade-offs they involve and when another option might be more suitable.

This should not encourage retailers to produce hundreds of thin pages targeting every possible AI prompt. The stronger approach is to improve the quality of information available around genuine buying decisions. Better comparisons, clearer specifications, useful buying guidance and original expertise improve the experience for human shoppers while also giving AI systems more evidence from which to construct useful answers.

A strong product page alone does not create complete trust. If an AI assistant quickly reduces thousands of options to three recommendations, customers may then want evidence that those recommendations deserve consideration. Reviews, reputable editorial coverage, independent product testing, expert recommendations and broader brand reputation can all help validate the shortlist.

This creates a growing overlap between AI visibility, digital PR and online reputation management. Brands should not only ask whether their own website contains enough information. They should consider what the wider web says about the product, which independent sources discuss it and whether those sources reinforce the claims being made by the brand. A company repeatedly describing itself as the best provides one type of signal; customers, journalists, reviewers and authoritative industry sources consistently recognising its strengths provides another.

The retailer website is therefore not disappearing from the journey. It is taking on an additional role. It remains somewhere humans evaluate products and complete purchases, but it can also become a source of information used by machines during discovery. The strongest ecommerce presence increasingly needs to serve both audiences without compromising the usefulness of the content for either.

For years, retail search teams have measured rankings, impressions, clicks, conversion rates and keyword demand. Those metrics remain important, but conversational discovery introduces questions they cannot answer. Does ChatGPT recommend your products for commercially important use cases? Does Gemini understand the difference between two products in your range? Which competitors consistently appear ahead of you? Are prices, specifications and availability represented accurately? Does the brand appear when shoppers ask broad category questions before they know its name?

These are prompt-level visibility questions, and they require a different type of benchmark. Retailers can begin by identifying the conversations customers are likely to have around discovery, recommendations, comparisons and product requirements, then testing how their own brand performs across those situations. The objective is not simply to generate a visibility score, but to understand where the brand enters or disappears from the customer journey.

If competitors dominate particular recommendation prompts, marketers can investigate what information or authority may be supporting them. If product details are regularly wrong, the priority may be data quality and consistency. If the brand only appears when explicitly named, the challenge may be broader category authority. AI visibility becomes much more useful when the result leads to a specific marketing action.

The most important implication from this research is not simply that 20% of UK consumers are beginning retail discovery with AI. It is that the point at which brands compete for attention is shifting. A customer's shortlist can increasingly be assembled before they visit a traditional search results page, marketplace or retailer website, which means part of the buying decision may already have happened before the first conventional click.

For marketers, the key question is increasingly straightforward: when an AI system is asked which products deserve consideration, is your brand included? If the answer is unknown, the starting point is measurement. Build the questions customers are likely to ask, identify which products and competitors appear, understand how accurately the brand is represented and then improve the signals supporting the most commercially important gaps. The shopping journey is not becoming entirely AI-driven, but AI is increasingly influencing which brands get invited into it.

The consumer figures referenced in this article come from research produced by World Retail Congress in association with Vercel, based on a survey of 4,000 European consumers examining how AI is reshaping shopping behaviour. UK-specific figures were reported by TechRadar, including the findings that 20% of UK consumers begin retail discovery using AI platforms, 87% find AI recommendations useful and 55% have purchased following an AI recommendation. OpenAI's merchant and shopping documentation was also used to support references to product feeds, merchant data and ChatGPT's shopping research experience.