Walmart's ChatGPT Checkout Flopped 3x
This hit Hacker News with 400 points today, and almost nobody is talking about the right part of it.
Walmart ran an experiment. They listed 200,000 products inside ChatGPT's Instant Checkout feature — letting people buy stuff without ever leaving the chat. The result?
In-chat purchases converted at one-third the rate of people who just clicked through to Walmart's website.
Let that sink in. A $400B company, with OpenAI's resources behind them, and the AI-native shopping experience was 3x worse than a normal website.
Why It Failed (And Why You Should Care)
Walmart's EVP of product and design called the experience "unsatisfying." That's corporate speak for "it sucked."
OpenAI has already confirmed they're phasing out Instant Checkout entirely. They're pivoting to app-based checkout handled by merchants instead.
So why did it fail? Three reasons that matter for indie hackers:
People Don't Trust AI With Money (Yet)
There's a psychological gap between chatting with an AI and actually giving it your credit card. When you're on Walmart's website, you already trust Walmart. When you're in ChatGPT, you're trusting a chatbot — and that trust hasn't been earned yet.
The conversion gap isn't about UX. It's about trust. And trust takes way longer to build than any of us want to admit.
Shopping Is a Browse Experience, Not a Query
When you're on a website, you can see 20 products at once. Compare prices. Read reviews. Scroll. Filter. That visual density is impossible in a text chat.
ChatGPT can describe a product, but it can't replace the experience of seeing it. Especially for physical goods — clothes, electronics, groceries. The chat interface strips away everything that makes shopping work.
The "Ooh, Add to Cart" Impulse Doesn't Exist in Chat
Walmart's website is designed to make you buy more. Related products. "Customers also bought." Deals in your face. That's not manipulation — it's good UX for commerce.
In ChatGPT, there's no impulse. You ask, it answers, you decide. There's no serendipity. No "oh, I also need that." The entire purchase funnel collapses.
What Walmart Is Doing Instead
They're not giving up on AI commerce — they're just admitting the chatbot-as-storefront model doesn't work. Their new approach: embed their own AI chatbot (called Sparky) inside ChatGPT.
Users log into Walmart. Carts sync across platforms. Checkout happens in Walmart's system, not OpenAI's.
In other words: use AI for discovery, but keep the transaction in your own environment. Smart.
The Lesson for Indie Hackers
Here's where this gets useful for you.
Don't build AI checkout. Build AI discovery.
If you're building an AI-powered commerce product, the data says people won't buy through a chatbot. But they will use an AI to find what they want — and then click through to a real site to complete the purchase.
Think about what that means for indie products:
- Affiliate content sites with AI search. Let people ask natural-language questions, get recommendations, click to buy. You get the commission without needing checkout at all.
- AI-powered comparison tools. "Compare the best standing desks under $500" — AI does the research, presents options, user clicks out to the merchant.
- Smart recommendation engines. Use AI to personalize product suggestions on your own site. Don't put the AI in front of the purchase — put it behind the browse.
The winning pattern isn't "AI replaces the website." It's "AI makes the website smarter." Keep the checkout. Let AI handle the discovery layer.
The Bigger Pattern
Walmart's failure isn't an AI failure. It's a context failure. They put a transactional experience into a conversational environment and wondered why it didn't work.
This is the same mistake every "AI will replace X" narrative makes. It assumes the interface is interchangeable. It's not. A website, a chat, a voice call — these are different contexts with different user behaviors.
The indie hackers who will win with AI aren't the ones trying to replace existing interfaces. They're the ones who understand which context fits which job.
Use AI to help people decide. Not to help them transact.
The transaction part — the actual checkout — let the user do that where they already trust the brand. Your AI adds value upstream, not at the point of sale.
Source: Search Engine Land — "Walmart: ChatGPT checkout converted 3x worse than website"
Want to Build AI Agents That Actually Work?
The OpenClaw Ultimate Setup gives you the complete framework for building autonomous AI agents — the kind that handle your workflows while you sleep. Pre-configured, tested, and ready to deploy.
$29 — Get the Ultimate Setup