AI shopping is good – what does that mean for retailers?

The move would be the latest development in what looks to be a fast disruption to the way consumers shop online. As I wrote about last week, the rate of visits to UK retailer websites from AI sources is growing by over 65% per month, according to Adobe data.

OpenAI and its reported partner Shopify did not publicly comment on the story about the checkout data.

In the last few months, search-enabled large language models have offered a novel shopping experience that is ultimately useful, fun and intuitive, even if there are some teething problems. OpenAI launched its shopping service through ChatGPT in April.

“We know online shopping can be overwhelming — but instead of juggling tabs or scrolling through endless results, you can just have a conversation. If you’re looking to do research about a specific category or product, ChatGPT gives you improved product results, product images, aggregated reviews from across the web and direct buy links to merchants. Ask follow-up questions, compare products, and tell ChatGPT exactly what matters to you all in natural language,” said an OpenAI spokesperson about the launch.

Soon after, Google announced the rollout of its AI Mode, which does a similar job through its proprietary AI Gemini 2.5, surfacing useful product information data on highly specific queries.The company is pulling on some of the masses of data it holds in servicing this product, such as the 50 billion product listings in its shopping graph, as well as product and place listings that can grab information on local inventory.

”Our AI tools help make shopping easier and more engaging, so people can connect with retailers and shop online with confidence. People are gravitating to AI-powered experiences, and AI features in Search enable people to ask even more questions in different ways, creating new opportunities for websites to be discovered,” said a Google spokesperson.

If OpenAI launched an integrated checkout, it would follow on from Perplexity, which allows its Pro subscribers in the US to check out in a chat window. It announced a partnership with payments provider PayPal in May and has been running a shopping service for subscribers in the US since last year.

Zero-click retail?

Consultancy firm Bain recently published a survey that showed 80% of search users in the US now rely on AI-generated summaries 40% of the time when using search. Their estimates suggest that this equates to a 15% to 25% dip in organic web traffic as users no longer feel the need to click through to websites. Media organisations like the MailOnline are already reporting massive drops in traffic, which it believes is explained by AI text summaries.

Google has pushed back on some of the coverage around traffic from search. According to Google, changing user preferences, algorithm updates and seasonal demand are all among the reasons site traffic may drop. Google Search’s vice president of product, Robby Stein, recently told The Economist that the company has not seen a dramatic decline in the number of outbound clicks, but it does not make this number public.

What differentiates retail from publishing is that the user typically wants more than just the information about the product; they also want to buy it. To do so, they currently need to navigate to a separate website. If AI websites start rolling out checkout functionality, that will no longer be the case for any retailer that chooses to partner with AI firms.

This begs the question: what should retailers do? Should they sign up to these partnerships and experiment with this new form of shopping if checkout features start being launched? Or should they hang back, given it remains unclear whether this will become a dominant shopping mode and, if it does, which LLM will be dominant in this space?

Retailers looking for a partnership should pick a provider that is easy to work with, but look for terms where they are not locked-in to only working with them in the future, says Catherine Brien, partner and managing director at AlixPartners. “You’re unlikely to want to lock yourself into one, but you probably do want to pick one to start with,” she says.

What can we learn from delivery and social?

There is a risk, says Mikey Vu, leader of Bain’s retail AI practice, that in failing to experiment in this space that retailers get left behind and lose prominence.

“This question popped up whenever a lot of these food and grocery and retail delivery services popped up as well,” he says. “There were many retailers who were very thoughtful about that and there were some retailers who didn’t believe that delivery was going to be a thing. Now they’re so far behind that their competitors have taken a lot of market share.”

Another close ecommerce parallel is the integration of checkout features into social media apps. But to give a sense of how condensed the timeline is: Instagram Checkout launched in 2019, nine years after the photo-sharing app was launched. Perplexity launched its search engine in 2022 and announced a one-click checkout just two years later.

Integrated tools like TikTok Shop have been growing fast, with Nielsen IQ reporting that it was the fastest-growing commerce platform in the UK in 2024. Brien says a lot of scenario modelling is required when working through a new channel like social. “In the instances where it’s been able to really grow brand and really grow incremental traffic, that is absolutely very powerful. Some brands have done that very successfully, but I’ve also seen retailers sort of crash and burn and cannibalise their own business,” she says.

“Beauty, fashion and low-ticket home categories have thrived in this space where impulse, education and influence drive conversion,” says Nicholas Found, head of commercial content at Retail Economics, about social. “But success hinges on discoverability, authority within a niche and the ability to adapt quickly to algorithm volatility.”

Shopify is already working with Perplexity to bring the brands on its platform to consumers shopping via the LLM-driven search engine. It is also creating a knowledge base app to help show ecommerce retailers what consumers are asking AI platforms about their brand. “As AI shopping agents continue to grow, we want to make sure merchants have what they need so their products are discoverable and their story shines through in every AI conversation. And we want this to happen by default where possible,” said Vanessa Lee, Shopify’s vice president of product.

Asked about direct partnerships with retailers, an OpenAI spokesperson said: “One of the things we’re exploring is an easy way for merchants to provide product feeds directly to ChatGPT, helping ensure more accurate, up-to-date listings. The best thing merchants can do at the moment is give us feedback on our website and sign up to get notified when feed submissions open.”

On how to get greater priority in AI-driven experiences, a retailer does not have to do anything differently according to Google. However, its recommendation is that high-quality product information is included in retailers’ merchant feeds and websites so that this can be pulled into relevant queries. 

Getting data back is a key ask

Another thing that retailers should be thinking about is what they get back from these firms as part of the deal. In addition to the seller fees, retailers need to consider the value of their first-party data. “Brands must demand visibility into what’s influencing recommendations – the equivalent of ad transparency in search. Access to first-party customer and behavioural data is critical. Without it, they’re flying blind,” says Found.

An interesting space to watch is the retailers with bucket loads of transaction data and the economic firepower to build their own AI rollout eventually. Amazon has already deployed conversational product discovery through its Rufus tool, while it has integrated conversational AI in its Alexa+ tool. Walmart has also announced plans for its own retail-specific LLMs.

Oz Ozturk, partner and AI lead for consulting at PwC, says there may be some bifurcation, with smaller retailers needing to seek out partnerships while larger retailers create their own Gen-AI-driven platforms. “It’ll be interesting to see what kind of monetisation and commercial models will come up,” he says.

The other key question is the format of AI search and the additional capabilities that suppliers are working on. Perplexity recently launched its own browser, Comet, with an LLM-driven sidebar assistant. The potential retail use cases for this format are fascinating: find yourself on a brand’s website looking at a specific TV, you could ask the sidebar assistant how highly the TV is rated or how much it costs at other retailers.

Also hovering in the background is the potential for AI agents to step forward and act on the product recommendations generated by the LLMs. OpenAI launched a ChatGPT agent last week and a rollout in Europe is coming soon. The OpenAI spokesperson pointed to an X post by Alex Banks, founder of AI newsletter Signal AI, where he used the tool to order a Tesco grocery shop including a sticky toffee pudding.

“This is so fast-moving and incredibly exciting,” says Brien. “I think it’s all to play for.”