More people are searching for their shopping using AI tools – here is what we know so far about how to get your brand featured
Key points:
- The number of UK visits to retail sites from AI search tools is growing at a rate of over 65% a month.
- Trusted content and information-dense product pages seem to be key.
- All of this is subject to change. Retailers need to think about tracking and adapting to changes in how both the models behave, as well as how consumers use them.
When shopping for products online, the first stop for many consumers is typing a product search into Google or whatever their favourite search engine is. That simple habit drives over £30bn in non-food retail spending per year, according to Retail Economics and Upp.ai, making it probably the most important way to discover products online.
Anybody looking at wider web analytics will know that something big is happening to that key part of the market. Adobe data for May shows UK visits to retail sites from AI sources grew by 66% month on month, 137% since the start of the year and 1,110% since the company started tracking the metric in August 2024.
This is all coming from a low base, but it is hard to avoid the thought we are living in a new epoch given ChatGPT is now the 5th most visited site in the world.
A survey of 2,000 consumers in the UK from February also commissioned by Adobe found that more than a third (35%) have tried out AI assistants when shopping online. “That level of adoption is pretty wild,” says Adobe lead analyst of digital insights Vivek Pandya.
Adobe is one of several companies traditionally tracking a brand’s web to start diving into the complicated subject of what large language models (LLMs) like ChatGPT and Google Gemini draw on when serving up answers. While we have long been focused on search engine optimisation (SEO), these businesses are saying we should now look at generative engine optimisation (GEO).
Edelman launched GEOsight in May, Ahrefs has Brand Radar while Semrush has its AI Toolkit. Much like with SEO trackers, this typically involves looking at how prominently a brand features in AI queries using certain keywords. Also being looked at is referral data, which shows the volume of traffic coming from the various LLMs.
“There are simple enough ways to set it up so you can track the traffic from the respective sites,” says PwC senior manager of strategy and AI go-to-market Daria Vlasova. “To understand where the demand is coming from is very important.”
Good SEO is still good GEO
But what makes a retailer stand out on AI search? It seems like a big factor remains how prominently they feature on traditional search engines.
That is not surprising given Google’s Gemini relies on Google and ChatGPT’s search, SearchGPT, which gets its information from Microsoft Bing. Yet the first thing that comes up as a search result is not necessarily what the AI will quote back to a user.
Recent analysis by Semrush found that ChatGPT regularly cited results that were on the third page of a web search, with Google’s Gemini and Anthropic’s Claude regularly doing the same. The slightly different behaviour of these models though already makes it a tricky challenge.
The launch of Google’s AI Mode in the US, where the search bar is replaced with the conversational interface we typically see in other LLMs, makes that even more difficult to work through.
“For the time being, at least the way to climb into AI Mode is to get on the first page of Google, which may not seem that insightful, but if we look at the fact that most of the citations that LLMs suggest are page three plus of Google, it’s counterintuitive,” says Semrush vice president of owned media Nick Eubanks.
What information do AI tools rely on?
Beyond that, understanding exactly what makes the LLM pick a certain product when you put in a shopping-related query, is not easy, but we have some ideas. First, it seems like the more information about the product and the shopping experience you have on your item pages, the better.
“What kind of return policy do you have? That could give you an edge versus someone else. How are you in shipping? How are you with your point system? Your loyalty programmes?” says Adobe’s Pandya.
Another key seems to be how heavily a product is cited by what the AI perceives as trusted sources. There have been various studies looking at this by different SEO providers looking at different models. A lot of it seems to vary, but the first key point is that user-moderated websites like Wikipedia, Reddit and Quora feature highly when it comes to indivdual domains. Also prominent are social media and video-sharing sites like Instagram and YouTube.
Another key source is legacy media, pages like Reuters, the New York Times and Forbes. The overriding takeaway here is that brands that tend to be part of a trusted conversation are the ones that the LLM is picking up.
This feels ironic, given the spats that AI companies are having with legacy media firms – a little bit more on that next week – but it seems like getting your brand in a major publication is somehow still a winning strategy despite the way that content consumption has moved away from publishers to social media.
Eubanks at Semrush says that he is redirecting his team’s content budget away from blog posts to focusing on updating old blog posts that are performing well and investing in distribution for the content they already have across social media sites and PR. “It’s like what a traditional marketing agency would do, but we’re doing it because we think this is where we have to go to get non-paid traffic.”
The final point is making your sites easily crawlable and the information easily grabbable. Other experts say this could include structured tables (for example, comparing your products to competitors), quick-loading pages, extensive product descriptions and embedded reviews.
“You need to start organising your site and your content in such a way that it is more easily digested by the AI engines that are out there,” says IBM partner and consumer centre of competency leader Karl Haller.
Everything is subject to change
So how much of this should retailers be doing? It seems quite clear that a lot of the above is just good online optimisation and marketing work. However, the big problem is that we are not at all sure how this is going to play out: Will these LLMs remain dominant? Or will shopping-specific platforms arise? Will the inner workings of the model change? We are all guessing at this stage.
However, what is clear, says Mikey Vu, who leads management consultancy firm Bain’s Retail AI practice, is that retailers need a framework in place about how to think about this as the different models develop. “That flexibility on the tech side is super important, because many big companies don’t have that agility to say oh my goodness, GPT just changed their model.”
1 Reader's comment