As the holiday shopping season approaches, both OpenAI and Perplexity have rolled out new AI-powered shopping features this week, embedding them into their chatbots to assist users in researching products before making purchases.
These new offerings are strikingly alike. OpenAI, for instance, suggests that users can ask ChatGPT to locate “a gaming laptop under $1000 with a screen larger than 15 inches,” or upload images of a luxury clothing item to find more affordable alternatives.
Perplexity, on the other hand, highlights how its chatbot’s memory can enhance shopping queries, allowing users to receive suggestions based on what the chatbot already knows about their location or profession.
Adobe has forecasted that AI-driven online shopping will surge by 520% this holiday, potentially benefiting AI shopping startups like Phia, Cherry, or Deft. However, with OpenAI and Perplexity expanding their AI shopping capabilities, these smaller companies may face new challenges.
Zach Hudson, who leads the interior design shopping platform Onton, believes that startups focusing on niche markets will still offer a superior experience compared to broader tools such as ChatGPT and Perplexity.
“The effectiveness of any model or knowledge graph depends on the quality of its data,” Hudson explained to TechCrunch. “Currently, ChatGPT and LLM-based platforms like Perplexity rely on search engines like Bing or Google, so their results are limited to what those indexes provide at the top.”
Julie Bornstein, CEO of Daydream and a veteran in e-commerce, shares this view. She told TechCrunch earlier this year that she always considered search to be “the overlooked aspect” of the fashion industry, as it rarely delivered satisfying results.
“Fashion is particularly complex and emotional — finding a dress you adore is a very different experience from shopping for a TV,” Bornstein said to TechCrunch on Tuesday. “Truly understanding fashion shopping requires specialized data and merchandising expertise that takes into account styles, materials, occasions, and how people assemble outfits over time.”
AI shopping startups often build their own databases, enabling their tools to be trained with more precise data — a task that’s more manageable when focusing on areas like fashion or furniture, rather than attempting to cover all possible topics.
For example, Hudson’s company Onton created a data pipeline to systematically organize hundreds of thousands of interior design products, which improved the training of its proprietary models. Still, Hudson warns that without such specialization, AI shopping startups risk being eclipsed by larger players.
“If a startup is only using generic LLMs and a chat interface, it’s tough to see how they can stand out against the tech giants,” Hudson remarked.
OpenAI and Perplexity, meanwhile, benefit from already having a large user base, and their scale allows them to secure partnerships with major retailers right away. While companies like Daydream and Phia send users to retailers’ sites to finish their purchases — sometimes earning affiliate commissions — OpenAI and Perplexity have integrated with Shopify and PayPal, respectively, so users can complete transactions directly within the chat.
These tech firms, which require vast and costly computing resources, are still searching for sustainable business models. Looking to Google and Amazon for inspiration, it’s logical for them to explore e-commerce — retailers might pay to have their products featured in search results.
However, this approach could end up worsening the very search problems that frustrate consumers today.
“Specialized models — whether in fashion, travel, or home goods — will deliver better results because they’re tailored to how people actually make decisions,” Bornstein commented.
Reporting support by Ivan Mehta.
