Photo: Canva
Will AI Take Online Product Search to Another Level?
A crop of AI-powered product discovery startups have arrived, including Lily AI, Pixyle.ai, Syte, ViSenze, and Vue.ai, to better match text and visual online product searches with what consumers are actually looking for.
Lily AI’s sophisticated algorithms, for instance, help link colloquial consumer searches typed in search bars — such as “quiet luxury,” “study hall,” or “boho chic” — to stock descriptions of goods, such as “midnight french terry active wear,” according to a profile in the New York Times.
The article notes that an estimated 70% of consumers quit their searches without making a purchase.
“It starts with understanding the language of the customer, and realizing just how poorly served they are in the online shopping experience today,” said Purva Gupta, Lily AI’s co-founder and CEO, in an interview with Center for Data Innovation. “Products are often put on shelves with legacy, ‘out-of-the-box’ attributes that come directly from manufacturers and distributors, and that don’t capture the nuance and detail that shoppers actually use when they’re looking for relevant products that match their needs.”
Macy’s, Bloomingdale’s, Gap Inc. brands, Abercrombie & Fitch, and thredUP are among Lily AI’s customers.
Pixyle.ai, which works with online marketplaces such as Depop and Otrium, offers a visual search tool that lets consumers upload an image of what they are looking for and get similar results. Pixyle.ai’s neural networks train its AI algorithms to not only identify but categorize fashion items by attribute, such as color or pattern, that match the keywords shoppers use when searching for an item.
Pixyle.ai notes in its marketing copy, “No words can describe an abstract pattern of a fashion item the same way an image can.”
The AI advances in product search come as Microsoft and Google both in February introduced ChatGPT-style chatbot products that led to speculation about how the technology would elevate online search.
In May, reports arrived that Amazon, which has long ranked as the top online product search channel, was looking to integrate generative AI into its search bar. In that same month, Google introduced an experimental product feature, Search Generative Experience (SGE), offering consumers relevant, up-to-date reviews, ratings, prices, and product images backed by Google’s Shopping Graph’s more than 35 billion product listings.
Elizabeth Reid, VP and GM of Google Search, said in a blog entry, “With generative AI in Search, we can help you understand the full picture when you’re shopping, making even the most considered and complex purchase decisions faster and much easier.”
Discussion Questions
Is product search on retail websites and marketplaces generally underwhelming, and will artificial intelligence improve results? Will text or visuals become the dominant form of online product search?
Poll
Loading …BrainTrust
Matthew Pavich
Sr. Director Retail Innovation at Revionics, an Aptos Company
James Tenser
Retail Tech Marketing Strategist | B2B Expert Storytelling™ Guru | President, VSN Media LLC
Search is one area where AI has the potential to enhance the user experience. There are many applications from allowing users to search for things in natural language to AI powering more tailored results. AI will also be able to learn and understand user preferences so that search can yield more personalized recommendations. That said, I see all of this as complementing, rather than completely replacing, traditional search functions. AI also raises concerns about bias in search results.
Anything that simplifies the search for products online is a welcome innovation and, in this scenario, AI is particularly well suited to make the experience of the shopper more pleasant.
AI will improve product search with more precise matches and recommendations of items often bought together. AI improves upselling, as shopper may search for a dress and AI will recommend matching shoes and handbags.
Visual search will become more dominant, as it’s faster than scrolling through product description pages.
I don’t think the retail industry yet recognizes the impact that AI search will have on their bottom line. Traditional search engines respond to queries by serving up a list of links to retailers, with adjacent ads that also link back to the corresponding retailer. With an AI search a shopper can describe in detail what they are looking for and be rewarded with very personalized and relevant suggestions — without any clickable links to the retailers, and without ads. This puts more power in the hands of the shopper to buy from whomever has the lowest price. It also bypasses the retail media and SEO revenue streams. It’s a whole new world.
I think the key will be how AI can help curate search results, rather than just expand the list I would come up with on my own. A longer list of links to click on won’t be helpful. A more focused list, curated to my criteria, will be hugely helpful. And I really like the way the article points out what AI can do with an image versus just text. To AI, a picture really will be worth a thousand words.
(Disclaimer – I work for Microsoft) I wrote a LinkedIn article in this arena back in May. While less focused on text vs image search it focused on how AI – and specifically Generative AI – could usher in new ‘aggregators’ who solve problems for consumers by working across different retailers.
Consider going out for dinner with friends – you may need a new outfit, a babysitter and something to feed the kids before you go out.
What if an AI concierge service could align all these for you across different providers while considering your overall preferences and profiles.
This would solve a clear ‘lifestyle problem’ – eliminating some hassle associated with organising family life.
This approach is not AI on its own – personalisation capability and clear perspective of the customer would still be needed but the approach is one many consumers may welcome.
Easier to compose more user friendly search aided by AI is a welcome advance but really only problem solves the tip of the iceberg. For most Retailers, supporting those attribute based, or image based searches are archaic Product Master Files (more than one usually) which have trouble supporting basic product descriptions, let alone boho, or quiet luxury. For Retailers with fast changing products (like boho) the odds of external sources (like social media) being able to support inferring the attributes in the 4 to 6 weeks that a product is current is not very high. Retailers carry a lot of technical debt in particular in the area of MDM/PIM which never seems to be addressed. A modern PIM may just become a competitive advantage vs a Retailer whose products never come to the top in search as even the best LLM can’t find values that don’t exist.
Sharp observation, Peter. Makes me speculate that those Product Master Files are due for an AI-powered re-creation.
Gen AI will definitely help customers search for things and allow algorithms to help consumers find things they weren’t even searching for. Coupled with other powerful AI tools, it will not only help the personalize and improve the shopping experience, but will also ‘lead’ consumers to products that help achieve business objectives as well whether those are more private label products, products with better sustainability scores, higher margin, etc. There is already a lot of proven value in the present as AI pricing has helped numerous retailers outperform the market – new advances will make AI even more formidable in the future.
Search and speed are critical functions that if are not optimized will absolutely cause visitors to abandon a site.
Making search more intuitive with AI sounds like a promising idea.
The present online product search experience is generally quite crappy, even on the best-functioning sites (like Amazon, Walmart, Home Depot). Users are challenged to string together the right words that zero in on the desired product. If items aren’t tagged with the right keywords, they can be virtually invisible to the shopper. Then their efforts are further thwarted by “suggested product” algorithms which crowd out valid results.
Adding image intelligence to the mix (“my desired product looks like this picture”) could be very helpful in some instances, fashion in particular.
Empowering shoppers with portable, personal AI agents that remember their preferences could be the real game changer.
There is huge potential in this space to curate the selection for the customer by utilizing AI. One particularly impressive feature is the ability of AI to refine results based on prior results. For example, I could be looking for an orange tweed shirt and on the first search, I am shown a variety of results – some wrong material/color, etc. From there, an AI search could enable the ability to select several of the results that I liked and say “find me more like this”. It would take it as a new search, but within the existing context of the initial one. Ultimately, the extra abilities that AI search methods are capable of can get the customer more of what they are looking for in a way that is tailored to their needs in the moment.
AI is important, as it has been for more than a decade on the back end. The question is willl “Generative AI” have a similar game changing effect?
Algos still largely drive what we end up buying online. Mainly b/c of the absence of better search and serendipity engines and tools. I think the combination of social data fed into algos, predictive AI and generative AI will herald a new age of “shopping” not just “buying” online.
Whether or not product search on retail sites is underwhelming largely depends on the technology the retailer is using. Legacy keyword-based technology struggles with understanding user intent and thus retrieving relevant results. Most product discovery solutions today layer AI overtop of these older engines, which doesn’t actually fix any of their limitations.
Next-generation AI-first search and product discovery platforms, however, utilize AI as their core and are far better at understanding user intent – and delivering relevant search results. In the near future, as more retailers adopt AI-first product discovery solutions, text will continue to be the dominant form of online product search as this is how customers are accustomed to searching. Getting them to change their behavior to something that has more steps, like visual search, is hard, and will only get harder as text searches get better thanks to AI.