Online search has been governed by a small handful of search engines for over 20 years, and the practice of search engine optimisation (SEO) has been a core part of every brand’s e-commerce marketing strategy since. But what happens when the entire system is upended by AI? Following a recent string of AI search updates from the biggest players in tech, it’s a question that’s at the forefront of brands’ and marketers’ minds, as they race to keep up with the tech developments that will influence how — and if — their products are found online.
“2025 is the year that search changed for the first time since the engines were invented,” says Matt Hildon, European retail portfolio director at Valtech, a digital agency that works with brands on their tech stacks and marketing strategies.
While Google still has a monopoly on online search, in April, its stock price dropped 7 per cent when Apple executive Eddy Cue testified that Google searches within Apple’s safari browser had declined for the first time since launch, during the Apple-Google antitrust trial. Cue linked this drop to the increasing use of AI chatbots for search, a thesis that is gaining traction thanks to several recent reports into changing user behaviour. Some 80 per cent of consumers now rely on AI-written results for at least 40 per cent of their searches, resulting in a 25 per cent drop in organic web traffic, according to research from consultancy Bain Co published in February. The research also found that 42 per cent of large language model (LLM) users now get shopping recommendations via AI platforms.
It’s a shift that the platform’s maker, OpenAI, is keen to exploit when it comes to commerce: the $300 billion tech company is piloting an integrated checkout product so that consumers don’t have to leave ChatGPT to complete purchases. It would charge brands a sales commission, according to a Financial Times report last week. OpenAI told Vogue Business it “cannot comment at this time.”
At the same time, Google is playing serious AI commerce catch-up: in May, it launched its “AI mode” for generative AI-led search, and announced three new AI shopping features (virtual try-on, Gemini-powered personalised shopping recommendations, and agentic price-match comparisons), which are all being rolled out in the US first. We can also expect an entry into AI search from Big Tech rival Apple soon, via its Safari browser, according to reports.
All this means ad agencies and marketing teams have shifted their focuses from SEO (search engine optimisation) to what’s been coined “AIO” (AI optimisation), and venture capitalists are backing a whole cottage industry of startups promising to increase brands’ AI search discoverability and conversions. While AI and marketing experts say it’s still difficult to know exactly what makes these LLMs tick, there are key website practices emerging that can help brands stay on the front foot of this shift.
When it comes to website content, context is king
The fundamental difference between LLMs and traditional online search engines is the way that users interact with them to find what they want. AI search has seen users shift away from pure keyword-based search prompts to longer, more conversational queries with follow-up questions.
At the same time, generative AI models are constantly upping the amount of context they can parse.
“As these models are used more, they process these longer prompts through their world model, which helps them understand the context behind the question that’s asked with all of the knowledge they’ve built up until now,” says Max Sinclair, CEO of Azoma AI, which helps brands with their AI search strategies. While Google has collected user data for 20 years — age, location, preferences — this will now all feed into AI search, which understands the searcher’s intent and context, based on this existing data.
Gone are the days of traditional keyword hacking, where brands would scope out the keywords that were easiest to rank on and build their website copy and product listings around. Instead, brands must now focus on writing intent-based content that talks about product categories and user problems more, as well as FAQ and how-to-style conversational content that pre-empts user queries. “A one-liner for a product description is no longer OK,” says Hildon. “Now, multiple pages of detail about a product can feed into the AI: the more you can expose it to data, the higher the chance of conversion.”
Data points that LLMs latch onto are details like defining who an item is for, talking about its functionalities, its use cases and what events a customer may wear it to.
For a luxury handbag, for example, a brand may have previously included a line in their product details such as “100 per cent Italian leather handbag.” Now, they may describe the product as a “100 per cent Italian leather handbag with space for a laptop that would suit a woman in her 20s or 30s, who can take it straight from work to an evening dinner”.
As well as curating more detailed product descriptions on their own websites, experts say that brands also need to think more holistically about every word relating to their brand that sits online.
“The question of brand health is now more relevant than ever, as LLMs are indexing the internet for all the unaided awareness and opinions about products and brands that exist,” says Sam Shapiro, partner at venture capital firm VMG Technology.
The more product imagery, the better
Another big shift with AI search is the move to “multimodal” search, where the question of context now extends to product imagery, as well as written content. Industry experts say the better a brand’s product images align with their written descriptors, the more likely they are to be parsed well and suggested to users by AI models.
“Three years ago, if you were to type ‘blue handbag’ into a search engine, you’d be served up red, brown and black bags, because the models didn’t understand the visual cues,” Sinclair says. This was due to brands’ practice of “keyword stuffing”, so that the bag would appear in as many search results as possible. That method will now work against brands, Sinclair adds. “The AI can tell if the colours match with the content and will prioritise for relevance: they’ll give the user a page of two or three much more relevant product recommendations, rather than a page of 50 bags in different colours.”
If a brand makes the bag in question in several colours, they will therefore need to show product images of every colourway on the listing page, to be surfaced in more AI searches.
But it’s not just a question of product images matching up with their descriptions when it comes to multimodal AI search. Google says that visual search through the company’s AI-driven image search Google Lens is its fastest-growing search behaviour, and that one in four of the 20 billion visual search queries it gets a month have commercial intent — for instance, users are searching to shop. Google says this change is being driven by younger consumers and Gen Z, whose visual, social and mobile-first experiences of the world is influencing the way they want to shop. Plus, the more angles of a product and detail that are shown in the product’s photo listings, the more likely it is to appear in the right visual search.
“When we speak to digital asset management software providers (DAMs), which sit in brands’ commerce software stacks, they say that video is now absolutely crucial for appearing more in multimodal search, too,” Hildon adds.
Stripping things back to basics: Sort out your code
Investing in adding AI-friendly detail to website content becomes futile if the infrastructure behind brands’ websites is difficult for LLMs to read. Brands might be unaware their websites’ code is blocking LLMs from crawling their site altogether.
“We’ve had a lot of clients who may have had attrition in their technical teams and are unaware that someone set up their website two years ago to block LLMs from scraping their content,” Sinclair says. “Once we’ve identified this, it would take about 30 seconds to change those lines of code, but often internal processes hold this transformation back.”
Sinclair stresses that brands’ tech teams need enough oversight so they can stay nimble and update websites in response to the latest AI developments.
“There’s a lot of fast technical hacks that tech teams can do — essentially web schemas’ AI upgrade — that enable them to understand your website better,” he says.
These include: LLms.txt files, which allow website text to be understood better by LLMs; robots.txt files, which instruct search engines on what pages they can access; and traditional structured data code (like a list of tags) that can be added to a website’s html, which still help search engines with AI capabilities parse a brand’s content better.
“Brands can win some pretty low-hanging fruit when it comes to a lot of the LLM crawling, and as with a lot of these more technical details, a little goes a very long way,” Sinclair adds.
As brands navigate all these changes, the consensus among experts is that AI search will ultimately better serve consumers. If brands can keep up with rapid tech advancements by providing the data on which AI models thrive, fashion e-commerce could become more relevant and more personalised than ever. And given that consumer opinions of brands are also fed into these LLMs, they point out that search’s new AI era could get a lot fairer.
“All sorts of brand health data is accessible to these answer engines,” says Shapiro. “Now there’s an extra element of discovery for brands’ healthy and quality, so ultimately I think that will allow better brands to bubble to the top — it’s the North Star for brands to show up that way.”
Sign up to receive the Vogue Business newsletter for the latest luxury news and insights, plus exclusive membership discounts.
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
