According to Wired, OpenAI’s ChatGPT already lets users buy Etsy products without leaving the app through Instant Checkout, while 60% of US consumers plan to use AI for shopping assistance and 20% would let AI handle everyday purchases completely. McKinsey projects up to $1 trillion in sales through agentic shopping by 2030 in the US alone, with OpenAI partnering with Walmart to enable ChatGPT purchases and both OpenAI and Perplexity striking deals with PayPal and Shopify. Google just introduced AI agents that can fill out checkout forms and call stores for pricing, but executives from seven tech companies reveal these features require significant user input, operate slowly, and only work for limited items due to ongoing negotiations about mistake prevention and data sharing.
Why your AI shopping assistant isn’t ready
Here’s the thing about these so-called agentic shopping experiences – they sound amazing in theory but fall apart in practice. I’ve tried several of these systems, and honestly? They’re clunky. You still need to babysit them through the entire process. The technology basically works like an overeager intern who needs constant supervision.
The real bottleneck isn’t the AI’s ability to understand what you want – it’s everything that comes after. These systems need to access product data, handle payment security, manage returns policies, and navigate the million edge cases that human shoppers deal with instinctively. And when something goes wrong? Who’s responsible when your AI buys the wrong size or misses a crucial discount?
The invisible data war
What most people don’t realize is there’s a massive negotiation happening behind the scenes about who shares what data. Retailers are understandably nervous about handing over their entire product catalogs and customer data to AI companies. Meanwhile, AI developers need that data to make their agents actually useful.
It’s a classic chicken-and-egg problem. The agents can’t get smart without data, but retailers won’t share data until the agents prove they’re reliable. According to a recent survey, consumer adoption ultimately depends on proven savings and protection guarantees that nobody can fully deliver yet.
Where we’re seeing progress
Now, it’s not all doom and gloom. Some prototypes are showing real promise. Expedia’s ChatGPT integration, for example, is apparently driving more sales than they expected, even though users still have to manually complete bookings. That suggests people find value in the discovery and comparison phase, even if the final purchase requires human intervention.
Google’s new form-filling agents and store-calling capabilities represent another step forward. Think about all the times you’ve abandoned a cart because the checkout process was too tedious. If AI can handle that grunt work reliably, that’s a genuine win. Recent data shows that 73% of shoppers are already using AI somewhere in their shopping journey, so the demand is clearly there.
So when will this actually work?
Looking at the agentic commerce projections and talking to people in the industry, I’d say we’re at least two holiday seasons away from truly hands-off AI shopping. The partnerships between Walmart and OpenAI and PayPal’s integration plans are laying crucial groundwork.
The reality is that shopping involves trust, and trust takes time to build. Would you really let an AI buy Christmas presents for your family without checking its work? Probably not this year. But as McKinsey’s analysis suggests, when these systems do mature, the impact could be massive. For now, think of AI as a shopping assistant rather than a replacement – it can help you find options and compare prices, but you’ll still need to make the final call.
