Google shakes up shopping with generative AI
By Maghan McDowell, Vogue Business
Google is doing for shopping what it did for search, adding a series of updates enabled by generative artificial intelligence. Google Shopping searches will now include AI-generated “briefs”, top product recommendations, a design-heavy personalised inspiration feed, and deal finding. The changes will be rolled out in the US starting today.
“The feeling in 2024 is similar to the shift in 2009 with mobile, but this year the shift is AI,” says Lilian Rincon, VP of product for Google Shopping. “It feels analogous to a special point in time.”
People who search for products on Google will now be served an AI-generated brief — similar to the platform’s new AI-generated search result summaries — using Google’s AI assistant Gemini to analyse the 45 billion product listings on Google. In addition to recommended products and products categorised based on various features, the brief will include information on what to consider when shopping for that type of product, based on content — including articles — found across the web. (Shoppers can click on the articles to learn more.)
The results will include the top three recommended products (based on sources from across the web), and an explanation of why they’re a fit. Sources initially will include user-generated content from sites such as YouTube, Reddit and Pinterest, Rincon says.
Users can narrow down results based on specific filters, such as size or a location. They can also submit more conversational queries, and the results will be more intuitive and assistive, according to Rincon. For example, someone searching for a dress for a wedding in Austin in November will be served options that take into account the weather in that city at that time of year. They could specify the price range and the results would be informed by past searches, so if the person normally searches for luxury brands, it would skew in that direction. Similarly, they could search for “secondhand”.
People who are signed into Google will also be served a photo-heavy, personalised Google Shopping home page with shoppable products and videos. This incorporates recent searches, what others are searching for across the web and people’s own preferences, such as favourite brands, which can be amended via “shopping preferences” on the Google Shopping menu. (Naturally, the more people share preferences with Google, the more Google also knows about how to serve personalised ads.)
Shoppers will also be served personalised deals, on top of existing price comparison and tracking tools. “Inspiration is fun, and for it to be fun, it needs to be personal,” says Rincon. She adds that while both the new approach to shopping search results and the personalised shopping feed will be personalised, the “assistive” nature of shopping searches will be less personalised. We will “take more leniency in telling you what we think”, she adds — more a tone of: “We want you to consider these things.”
In a blog post announcing the updates, Google’s VP and general manager of consumer shopping Sean Scott called it a “reinvented” Google Shopping that is “rebuilt from the ground up”. (He also noted that there are more than a billion searches via Google Shopping every day.)
The challenges of applying this to fashion
Fashion and luxury searches come with unique challenges, including understanding taste and price point, avoiding dupes, and what to base the brief on; Reddit reviews are not often a go-to source for fashion shoppers. For now, the AI-generated shopping briefs will be labelled as “experimental”, and people are invited to provide feedback.
In a demonstration for Vogue Business, the search produced results for both new and secondhand luxury products without being asked to. “We are still learning,” Rincon says, adding that Google is working to make apparel searches more personalised on both search and the personalised feed. She notes that Google has tested a “yes or no” style-rating tool, in acknowledgment that “style and aesthetic categories are becoming more nuanced”. The aim is for the new shopping search approach to offer results that are “much smarter” than those from purely using the type of structured data (like taxonomy and colour) that is typical of e-commerce searches.
Some fashion and beauty retailers have been updating their own search tools with AI, adding the option for people to search more conversationally to receive more intuitive results that lean on generative AI and computer vision, rather than being limited to taxonomic searches that are limited to specific, exactly matching search terms. This is especially true for those with large and varied catalogues. Resale site Vestiaire Collective just overhauled its search capabilities, while rental fashion platform Rent the Runway is currently testing AI search. Revolve just updated and improved its search capabilities with AI. Amazon, similar to Google, is also using AI to provide automated summaries of product feedback.
Long-time retail executive Julie Bornstein just raised $50 million to build an AI fashion search platform. Her previous fashion search startup The Yes was acquired by Pinterest in 2022, which has long hung its hat on the understanding that pictures are often more accurate than words when it comes to matters of style and taste.
To adapt to the new Google Shopping search, brands are encouraged to make sure their product data is up to date and accurate, with reviews and images being particularly important, Rincon says.
Google has added other AI shopping updates. In June, it began enabling people shopping for women’s shirts to see the item on a wide range of models, later adding dresses to the options. It also enables people to upload product images to source similar shoppable products and to use text prompts to generate dream product images that it can then match across the web.
In May, it began rolling out “AI overviews” as part of any Google search result, which includes AI summaries of information and links out to sources. The occasionally inaccurate nature of the results — such as the now-well-known suggestion to add glue to pizza toppings — is a reminder of how experimental this science is.
As people use the new Google Shopping tools, the aim is for results to become more accurate and useful, Rincon says. “We are still just getting started.”