As the holiday shopping season approaches, OpenAI and Perplexity have unveiled new AI-powered shopping assistants integrated into their chatbots, enabling users to find products through natural language queries. OpenAI’s ChatGPT can now help users search for items like a gaming laptop under $1,000 with a screen larger than 15 inches, or analyze a photo of a designer outfit and suggest similar, more affordable alternatives. Perplexity highlights its chatbot’s memory feature, allowing it to tailor recommendations based on a user’s location, job, or past preferences. Despite the excitement around these tools, startups focused on niche AI shopping experiences aren’t panicking. Experts argue that specialized platforms still hold a significant edge over general-purpose AI assistants. Zach Hudson, CEO of Onton, a design-focused shopping tool, points out that the quality of any AI model depends entirely on its data sources. “Right now, ChatGPT and Perplexity rely on existing search indexes like Bing or Google,” he said. “That means they’re only as good as the top few results those systems return.” This limitation hampers their ability to deliver deep, context-aware recommendations, especially in complex categories like fashion or home decor. Julie Bornstein, CEO of Daydream and a veteran in e-commerce, echoed this sentiment, calling search “the forgotten child” of the fashion industry. “Finding a dress you love isn’t like finding a TV,” she said. “It’s emotional, nuanced, and involves silhouettes, fabrics, occasions, and how people build entire wardrobes over time.” These subtleties require domain-specific knowledge that general AI models struggle to replicate. AI shopping startups like Onton, Phia, Cherry, and Deft have built their own curated datasets and internal models trained on high-quality, category-specific information. Onton, for example, developed a dedicated data pipeline to organize hundreds of thousands of interior design products with precision, enabling its AI to understand design trends and spatial relationships better than off-the-shelf models. Hudson believes that startups relying solely on generic large language models and conversational interfaces won’t be able to compete with giants like OpenAI and Perplexity unless they invest heavily in specialized data. “If you’re just using off-the-shelf LLMs and a chat interface, it’s hard to see how you can win,” he said. However, OpenAI and Perplexity have key advantages. They already have massive user bases, and their scale allows them to secure direct partnerships with major retailers. OpenAI has teamed up with Shopify, while Perplexity has a deal with PayPal, enabling users to complete purchases directly within the chat interface. In contrast, most AI shopping startups still redirect users to retailers’ websites, often earning revenue through affiliate programs. While these big players face challenges in achieving profitability due to the high cost of running AI systems, e-commerce could be a viable path forward—potentially through product advertising in search results. But Bornstein warns that if large platforms prioritize monetization over user experience, they risk worsening the very problems consumers already face with traditional search. Ultimately, experts agree that vertical AI models—those trained specifically for fashion, home goods, or travel—will outperform general models because they’re built around real consumer decision-making processes. Specialization, not scale, may be the true differentiator in the future of AI shopping.