OpenAI scales back ChatGPT shopping plans as Instant Checkout feature fails to gain traction

OpenAI's Pivot Into Shopping Has Been a Disaster

Can't-miss innovations from the bleeding edge of science and tech OpenAI's efforts to transform online shopping are faltering. According to new reporting from The Information, the company is walking back its plan to allow users to buy products suggested by ChatGPT directly inside the chatbot. Now, the company will route users to a connected third-party app, where they can input payment information and finalize the purchase. "We are evolving our commerce strategy within ChatGPT to better meet merchants and users where they are," an OpenAI spokesperson told the outlet. "Instant checkout is transitioning to apps, where purchases can occur more seamlessly." It's a small change that betrays a major shift in OpenAI's approach to online shopping. Last September, it launched "Instant Checkout" for ChatGPT, which allowed users to browse products from selected retailers and complete purchases without leaving the chat window. The feature was launched in partnership with Shopify, Etsy, and other large shopping platforms like Walmart and Target followed, in a show of the effort's ambition. This, in theory, posed an existential threat to retailers that didn't get on board. People are increasingly using AI chatbots to browse the web, look up information, and indeed find products, making it less likely for someone to directly access a brand's website. If ChatGPT's anaconda grip around the web tightened, then retailers faced getting iced out of its ecosystem, meaning fewer sales. But OpenAI's efforts have hit a few reality checks. One, according to The Information's reporting, was that OpenAI's data showed few users were finalizing their purchases inside the chatbot, despite many of them using it to browse for products. This gave OpenAI little incentive to address the other big issue: that acting as a storefront requires a lot of hard work. Providing live prices and other data on millions of products from countless merchants is an enormous undertaking. If ChatGPT provides data that's slightly inaccurate or out of date, it could cause a transaction to fail. Then there's the additional responsibility of dealing with refunds, cancellations, and fraud prevention, all while complying with tax and consumer protection laws. News of OpenAI's shopping scaleback sent shares of Expedia and Tripadvisor soaring by 8 percent and 13 percent respectively on Thursday. Investors had feared that AI agents could cut online travel agents out as middlemen to make bookings and travel itineraries. While it's cause for celebration for those companies, this isn't OpenAI admitting defeat on online commerce. Many of its purported 700 million actively weekly users still use ChatGPT for product recommendations. But it's a clear setback in its push to let ChatGPT act as an omnipotent department store, which could have implications for how other AI companies' efforts in this area fare. Competition is heating up: Meta is testing its own AI shopping research tool to rival OpenAI's, Bloomberg reported, which currently doesn't offer a checkout or payment option within its chatbot. In a note quoted by Business Insider, analysts at TD Cowen called OpenAI's reversal a "stunning admission." "The news signals that AI platforms replacing apps to become the 'new OS' is either not playing out, or at a minimum is pushed back significantly," the Thursday note to investors read.

Agentic Commerce in Doldrums? OpenAI Scales Back Shopping Plans for ChatGPT

Maybe it was a case of over-estimation of the market or maybe just a case of keeping the focus on the right things like GPT-5.4 In a week where news headlines were dominated by quarterly results and the US government's targeting of an AI startup and replacing it with a more compliant one, an announcement from OpenAI and its impact on one specific group of stocks was missed out. OpenAI announced a retreat from offering direct shopping via ChatGPT and the result was a resurgence in stock prices of travel and food delivery firms. A report by Forrester quoted one from The Information to suggest that OpenAI had nixed the "instant checkout" functionality that enabled shoppers to fulfil their orders directly from their ChatGPT window. Of course, there is no official word from the AI startup, that recently saw a fresh fund infusion at an all-time high valuation of $840 billion. Per the report, OpenAI reportedly scaled back on the feature as a result of which the checkout feature would move back into the merchant apps within ChatGPT or be directed back to their websites, as has been the case for several years in the search engine era. So what exactly was Sam Altman hoping to change? For years shoppers searched and went to the relevant websites to complete their orders. However, OpenAI had us believe that customers will ask questions about products to ChatGPT or Perplexity and buy it right there within the thread, never ever visiting a retail website again. They announced the era of agentic commerce. Quite honestly, the hype was built around hope rather than data. In fact, a research from Forrester actually noted that Gen Z used social channels more than answer engines (chatbots) for product discovery. It said less than a quarter of them in the US used ChatGPT to search for products while for older generations it is even less. Though the news cycles did not quite respond to OpenAI's shifting stance, the stock markets did. On a day when the S&P 500 fell by 0.5%, a group of stocks made up of travel-booking and food delivery firms bucked the trend. Companies like Booking Holdings, Expedia and DoorDash spiked during the trading time and after hours. In recent times, many of these companies faced selling pressure over fears of how AI could potentially disrupt the business. The hype created by OpenAI over agentic commerce caused investors to worry that chatbots may take away or definitely undercut consumer apps. The issue here isn't about lifting some stocks in a few trading sessions. It is about how OpenAI is doing the fox-trot with its strategy and product development. Instant Checkout was launched with a lot of fanfare five months ago with the promise that they could buy stuff "without ever leaving the chat", but now they are back to sending the shoppers to merchant apps. One-step-forward-two-steps-back? Maybe not, because ChatGPT still gets users through the journey of discovery to fulfilment within a couple of clicks, compared to what its competitor Gemini still does. Makes us wonder why OpenAI has reversed a decision without even giving it time to explore and explode. Isn't it rather obvious that user mindsets wouldn't change that quickly towards discovery and that it needs to be built on long-term trust? The decision coincides with OpenAI launching its Thinking GPT-5.4 Maybe they didn't want to spread themselves too thin. Last week, the company also released its latest foundation model - the GPT-5.4 which has been billed as their "most capable and efficient frontier model for professional work (our bold)." Maybe Altman is hoping to consolidate with the $200 million he earned by displacing Anthropic from the US government and target more enterprise business - which today has emerged as the sole path towards profitability. By the way, OpenAI is claiming that the new model will improve token efficiencies and solve the same problems with significantly fewer tokens than its predecessors. It also claims significantly improved benchmark results. OpenAI has also reworked the API version of GPT-5.4 with regards to managing tool calling through a new Tool Search system. In the earlier models, the system prompts would list definitions for all available tools that resulted in more token consumption than was warranted. The new system allows models to seek out tool definitions as required that makes the outcomes faster and cheaper. However, the most important and some would say critical shift has been the introduction of a new safety evaluation to test the chain-of-thought of new models. You can read what OpenAI has to say on this front right here though from the looks of it, OpenAI is hoping that the new system would reduce deception levels and GPT-5.4 will be the Thinking Version that everyone, including OpenAI detractors, were waiting for.