As the holiday shopping season approaches, AI giants OpenAI and Perplexity have both announced this week the launch of AI shopping features, integrating them into their existing chatbots to help users research potential purchases, officially entering the trillion-dollar e-commerce market.

OpenAI suggests that users can ask ChatGPT questions such as "looking for a new laptop under $1000 with a screen over 15 inches and suitable for gaming," or ask via images if there are cheaper similar high-end clothing options. At the same time, Perplexity emphasizes its chatbot's memory function, which can provide personalized shopping recommendations based on user-known information, such as location or job.

AI E-commerce Robot Artificial Intelligence (1)

Market Forecasts and Challenges for Startups

Adobe predicts that AI-assisted online shopping during this year's holiday season will grow by 520%. This should have been a blessing for AI shopping startups like Phia, Cherry, or Onton (formerly Deft). However, with the entry of OpenAI and Perplexity, these vertical-focused startups face significant competitive pressure.

The core advantage of OpenAI and Perplexity lies in their large user base and established partnerships with retail giants. Unlike Daydream or Phia, which redirect users to retailers' websites for purchases (earning through affiliate marketing), OpenAI has partnered with Shopify, while Perplexity has partnered with PayPal, allowing users to checkout directly within the conversational interface. Additionally, treating e-commerce as a revenue stream, monetizing through retailer-paid ads also aligns with the business logic of these companies to cover their substantial computing costs.

Vertical Markets: Specialization is Key to Survival

Despite the strong presence of general AI tools, executives of AI shopping startups focusing on specific fields remain optimistic about their advantages.

Zach Hudson, CEO of interior design shopping tool Onton, told TechCrunch that the quality of general AI models depends on their data sources. Currently, tools like ChatGPT and Perplexity still rely on existing search engine indexes like Bing or Google. He believes that vertical-focused startups can offer a better user experience through specialized data. For example, Onton has developed dedicated data pipelines to categorize hundreds of thousands of interior design products in a clearer way, training internal models with high-quality data.

Julie Bornstein, a veteran e-commerce executive and CEO of Daydream, also agrees with this view, especially in the fashion sector. She pointed out, "Fashion... has unique subtleties and emotions—finding a dress you love is entirely different from finding a television." She believes that a deep understanding of fashion shopping comes from specific domain data and product logic, capable of grasping the silhouette, fabric, occasion, and matching of clothes.

Both executives emphasize that if startups only use existing LLMs and conversation interfaces, they will find it difficult to compete with big companies. "A vertical industry model—whether in fashion, travel, or home goods—will perform well because they are closer to consumers' real decision-making processes