OpenAI and Perplexity roll out AI shopping tools as niche rivals stress value of specialisation

OpenAI and Perplexity have unveiled new AI-powered shopping assistants ahead of the holiday season, integrating retail-focused features into their existing chat platforms to help users research and refine purchase decisions.

Both tools function in similar ways. OpenAI said users can ask ChatGPT to identify items such as a gaming laptop under $1,000 with a screen larger than 15 inches, or upload photos of high-end clothing to receive suggestions for more affordable alternatives. Perplexity is leaning on its chatbot’s memory to personalise shopping searches, stating that users can request recommendations tailored to details the system already knows about them, including location or occupation, as reported by TechCrunch.

Adobe has forecast a 520% surge in AI-assisted online shopping this festive period — an opportunity for specialist shopping startups such as Phia, Cherry, and Deft (recently renamed Onton). Yet the rapid expansion of OpenAI and Perplexity into this space raises questions about how smaller players will compete.

Zach Hudson, chief executive of Onton, told TechCrunch that sector-specific AI remains more capable than broad, general-purpose models. He stated that any model or knowledge graph is only as strong as its underlying data sources, noting that ChatGPT and Perplexity rely on search indexes such as Bing or Google, limiting them to the quality of results those engines provide.

Daydream chief executive and long-time e-commerce leader Julie Bornstein echoed this view and spoke to TechCrunch. She remarked that search has long been an underdeveloped part of the fashion industry, observing that fashion shopping is highly nuanced and emotional, requiring a depth of understanding that general models struggle to match. She stated that identifying a preferred dress is fundamentally different from choosing a television, and argued that domain-specific datasets and merchandising logic — spanning silhouettes, fabrics, occasions and outfit-building — are essential.

AI shopping startups often build their own bespoke datasets, training their models on cleaner, higher-quality information tailored to specific sectors. Onton, for instance, developed an extensive pipeline to catalogue hundreds of thousands of interior design products more effectively. Hudson warned that without such specialisation, startups risk being overtaken, as per TechCrunch.

He argued that relying solely on off-the-shelf large language models and conversational interfaces leaves little room to compete with major tech companies.

OpenAI and Perplexity, however, benefit from their existing user bases and their ability to secure partnerships with major retailers. While players such as Daydream and Phia redirect shoppers to external retailer sites — sometimes earning affiliate commissions — OpenAI has integrated Shopify, and Perplexity has linked with PayPal, enabling users to complete purchases directly within the chat interface.

Both companies continue to search for sustainable business models, given the substantial computing costs involved in running large language models. Following the precedents of Google and Amazon, they may look to e-commerce revenue streams, including retailer-paid product placement within search results.

Bornstein suggested that such an approach risks amplifying the frustrations consumers already experience with search, and maintained that vertical, domain-specific models — whether in fashion, travel, or home goods — will ultimately deliver superior performance because they are aligned with real customer decision-making.