Why ChatGPT Checkout Failed (And What Perplexity Got Right)

ChatGPT's Instant Checkout launched with enormous expectations and landed with a 1.18% conversion rate and 77.45% cart abandonment. OpenAI tried to build a full shopping experience inside a chatbot. It did not work. Meanwhile, Perplexity took the opposite approach -- reduce friction with existing payment rails -- and expanded to free users while OpenAI scaled back. For Shopify merchants, the lesson is clear: data quality and checkout simplicity matter more than AI sophistication.

The promise of AI shopping -- and why merchants paid attention

Through 2024 and early 2025, AI shopping was positioned as the next major channel. OpenAI, Google, Perplexity, and Amazon raced to build experiences that browse, compare, and buy on behalf of consumers. The pitch to merchants was compelling: a new sales channel where AI agents recommend your products to millions of users, no ad spend required.

Shopify merchants had special reason to watch. Shopify's Agentic Storefronts feature promised to sync product catalogs directly to AI platforms, making your inventory discoverable by ChatGPT, Perplexity, and Copilot. The implication was unmistakable -- if you had your structured data in order, AI would become a traffic source that rivals paid search.

The market numbers reinforced the urgency. The agentic commerce market is projected to hit $60 billion in 2026, growing at a 45% CAGR. ChatGPT alone processes roughly 50 million shopping-related queries daily. 61% of merchants remain unprepared for AI commerce. Tools like Analytics Agent's AI ranking tracker help merchants monitor where they appear in AI shopping results -- but visibility tracking cannot fix a broken checkout.

AI shopping was always going to matter. The question was who would build the right version of it. OpenAI's first attempt offers a case study in what happens when you get the fundamentals wrong.

What OpenAI built -- from Operator to Instant Checkout

OpenAI's approach to AI commerce unfolded in three phases, each a response to the failures of the previous one.

Phase 1: Operator (January 2025)

In January 2025, OpenAI launched Operator -- an AI agent that could browse the web, fill forms, and complete purchases on third-party sites autonomously. The vision was ambitious: a general-purpose shopping agent that navigated real websites like a human would. Operator could open a browser, search for products, add items to carts, and enter payment information.

The problems were immediate. Operator was slow, clumsy with CAPTCHAs and authentication flows, and frequently hit errors on checkout pages that required account creation. Merchants had no way to optimize for it, and conversion rates were dismal. The user experience resembled watching someone buy something online for the first time -- slow, error-prone, and frustrating.

Phase 2: Deprecation (August 2025)

By August 2025, OpenAI deprecated Operator. The general-purpose web-browsing approach was too unreliable for ecommerce. Too many checkout flows were unique, too many sites required authentication, and too many product listings had stale data. OpenAI effectively admitted that trying to navigate the open web for commerce was a dead end.

Phase 3: Instant Checkout (late 2025)

OpenAI pivoted to ChatGPT Instant Checkout -- a native, in-conversation purchase experience. Instead of browsing external sites, Instant Checkout pulled product data into ChatGPT and attempted to handle the entire transaction within the chat interface. This was the approach that was supposed to fix everything Operator got wrong.

It did not.

The numbers: 1.18% conversion and 77% abandonment

CNBC reported what many in the industry suspected: "OpenAI's First Crack at Shopping Stumbled." The data paints a stark picture.

ChatGPT Instant Checkout performance:

For context, the average ecommerce conversion rate sits between 2.5% and 3.0%. Instant Checkout underperformed the industry average by more than half. And a 77.45% abandonment rate, while not far from the industry average of roughly 70%, is devastating when combined with such a low starting conversion rate. Very few users were adding items to begin with, and most of those who did walked away.

These are not numbers that suggest a product in early growth. They suggest a product with fundamental structural problems.

If you are a merchant trying to understand how AI platforms convert traffic to sales, the AI shopping agent benchmarks guide provides a detailed comparison across ChatGPT, Perplexity, Amazon, and other platforms.

Three reasons ChatGPT checkout failed

The failure was not about AI sophistication. OpenAI has some of the most capable language models on the planet. The failure was about commerce fundamentals that no amount of AI can paper over.

1. Stale and inaccurate product data

Instant Checkout pulled product information from sources that were not updated in real time. Prices were wrong. Products showed as available when they were out of stock. Delivery estimates did not reflect actual shipping times or costs.

For a consumer, there is nothing more damaging to trust than seeing a price in ChatGPT and a different price at checkout. Or adding a product that turns out to be unavailable. Every inaccuracy is a reason to abandon.

This is the same data quality problem that plagues merchants in analytics. If your product data is not accurate, clean, and current, every downstream system breaks -- whether that system is GA4, Google Shopping, or ChatGPT's checkout. For merchants tracking how AI platforms interact with their catalogs, monitoring ChatGPT traffic in GA4 is the first step toward understanding what AI shoppers actually see.

2. Checkout friction inside a chatbot

Completing a purchase requires trust signals that chatbots struggle to provide. Consumers expect to see the merchant's branding, their return policy, customer reviews, and a familiar payment form. Instant Checkout replaced all of that with a generic, AI-mediated interface.

The user had to trust that OpenAI was accurately representing the product, the price, and the merchant. There was no merchant identity in the checkout flow. No social proof. No store-specific return policy visible at the moment of purchase. The checkout was optimized for AI convenience, not human confidence.

Compare this to buying on a merchant's actual Shopify store, where the brand, reviews, policies, and payment options are all visible. Or buying through Shopify's Shop app, where the checkout infrastructure is battle-tested. Removing the merchant from the checkout removed the trust that makes people pull out their credit cards.

3. Limited product selection and discovery

Instant Checkout did not have access to the full breadth of ecommerce. The catalog was limited, skewing toward commodity products where price and availability are the primary differentiators. For merchants selling unique, branded, or differentiated products, ChatGPT's shopping experience simply did not represent them well.

Discovery was also a problem. The recommendation algorithm favored products with the most structured data available -- but much of that data was incomplete or inconsistent. Merchants who had invested in complete JSON-LD schema and structured data had a visibility advantage, but even that could not overcome the broader platform issues.

What Perplexity did differently

While OpenAI was building and then tearing down its own commerce infrastructure, Perplexity took a fundamentally different approach. And it worked.

Friction reduction, not friction creation

Perplexity's shopping model does not try to replace the checkout. It enhances the discovery phase and then hands off to trusted payment rails. Perplexity integrated with PayPal, enabling users to pay with saved payment information in a single step. The merchant still fulfills the order. The payment still flows through established infrastructure.

This is friction reduction, not friction replacement. Perplexity did not ask users to trust a new checkout system. It plugged into one they already trusted.

Real-time data and transparent sourcing

Perplexity shopping features pull data from live merchant sites and clearly cite their sources. When Perplexity recommends a product, the user sees where the information comes from. Prices and availability are pulled from the merchant's actual storefront, not a stale index.

This transparency builds trust. The user knows they are looking at current data from a real store. It is the opposite of Instant Checkout's opaque product cards where the data source was unclear and accuracy was hit-or-miss.

Expanding access instead of pulling back

While OpenAI was scaling back its shopping ambitions and pivoting to retailer-specific apps, Perplexity expanded shopping features to its free tier. More users getting access to a simpler, higher-trust shopping experience. Perplexity bet on volume through a superior user experience. OpenAI bet on sophistication through a technically ambitious but fundamentally flawed one.

For Shopify merchants tracking referral traffic from AI platforms, the difference is visible in the data. Perplexity referral sessions tend to convert at higher rates than ChatGPT shopping sessions. Understanding these patterns starts with tracking Perplexity referral traffic to your Shopify store.

Action: Already seeing AI referral traffic? Analytics Agent's LLM Traffic Dashboard breaks down sessions, conversions, and revenue by AI platform -- so you can compare ChatGPT and Perplexity performance side by side.

The pivot: from checkout to retailer apps

After the Instant Checkout stumble, OpenAI shifted strategy. Instead of building a universal AI checkout, they pivoted toward retailer-specific apps integrated into ChatGPT. Walmart launched Sparky, its ChatGPT app. Etsy built a dedicated integration. The bet is that retail brands with established trust, fulfillment infrastructure, and customer relationships are better positioned to own the checkout -- even inside an AI interface.

This pivot is telling. It acknowledges that the checkout is the retailer's domain, not the AI platform's. The AI's job is discovery and recommendation. The merchant's job is conversion and fulfillment.

For Shopify merchants, this is significant. Shopify's Agentic Storefronts are positioning your store to be the integration layer between AI platforms and your customers. Rather than handing your checkout to ChatGPT, you maintain your storefront, your brand, and your customer relationship while AI platforms drive discovery.

The question is whether Shopify will build its own ChatGPT apps or whether individual merchants will need to. Right now, the approach is catalog syndication through Agentic Storefronts -- your products show up in AI results, but the transaction stays on your infrastructure.

Action: Audit your Shopify store's readiness for AI shopping agents. Analytics Agent's JSON-LD audit validates that your product schema, pricing, and availability data are complete and accurate -- the foundation that AI agents need to recommend your products correctly.

What this means for your AI visibility

The ChatGPT checkout failure does not mean AI shopping is dead. It means the first attempt at full-stack AI commerce was premature. The underlying trend -- consumers using AI to discover and evaluate products -- is accelerating.

What changes is the model. Instead of AI platforms owning the entire funnel from discovery to checkout, the winning model looks like AI platforms owning discovery and merchants owning conversion. That makes your AI visibility -- whether AI platforms can find, recommend, and accurately represent your products -- more important than ever.

Analytics Agent for Shopify tracks this across multiple dimensions. The LLM Traffic Dashboard shows sessions, conversions, and revenue from ChatGPT, Perplexity, Claude, Gemini, and other AI platforms. The AI Ranking Tracker monitors whether your brand appears in AI shopping recommendations for your target queries. Together, they tell you whether AI agents can find you and whether that discovery converts to revenue.

The merchants who win in agentic commerce will be the ones who can measure their AI visibility, maintain clean product data, and prepare their stores for AI shopping agents before the next wave of platform features rolls out.

How to win in agentic commerce: lessons from the ChatGPT checkout failure

The ChatGPT Instant Checkout failure offers concrete, actionable lessons for every Shopify merchant. These are not speculative predictions. They are principles validated by what actually happened when the most well-funded AI company in the world tried to build a checkout.

1. Data quality beats AI sophistication

OpenAI has GPT-4, massive compute, and 200+ million weekly users. None of that mattered because the product data was stale and inaccurate. Your product schema -- prices, availability, descriptions, images, shipping details -- is the API that AI agents use to represent your products. If your data is wrong, the AI's recommendation is wrong, and the customer leaves.

What to do: Run a structured data audit on your store. Ensure every product has complete, accurate JSON-LD markup including price, priceCurrency, availability, brand, description, and image. Update this data whenever prices or inventory change. If you are on Shopify, your Liquid templates should pull directly from live product data -- never hardcode prices or availability in schema.

2. Own your checkout -- do not outsource it

The merchants who will thrive in AI commerce are the ones who maintain their checkout experience. Your checkout is where trust converts to revenue. Your branding, your return policy, your reviews, your payment options -- these are competitive advantages that no AI intermediary can replicate.

What to do: Optimize your Shopify checkout for conversion. If you are still on checkout.liquid, migrate to checkout extensibility before the August 2026 deadline. Ensure your checkout loads fast, displays trust badges, and supports multiple payment methods including Shop Pay, PayPal, and Apple Pay.

3. Measure AI as a channel, not a curiosity

ChatGPT sends real traffic that produces real revenue. Perplexity does too. These are not novelties -- they are measurable acquisition channels. If you are not tracking AI referral traffic separately from generic referral or direct traffic in GA4, you are making decisions without data.

What to do: Set up AI referral tracking in GA4. Identify traffic from chatgpt.com, perplexity.ai, and other AI domains. Track conversion rates by AI platform. Compare AI channel performance to organic search and paid acquisition. Analytics Agent's LLM Traffic Dashboard does this automatically across six AI platforms.

4. Invest in structured data as your AI storefront

The merchants who appeared in ChatGPT shopping results were the ones with the most complete product data. Structured data is not just an SEO tactic anymore -- it is the foundation of AI discoverability. Every AI shopping agent evaluates products using schema markup, reviews, and pricing data.

What to do: Treat your JSON-LD schema as a product data feed for AI. Audit it for completeness. Add AggregateRating, Review, Brand, and Offer schema to every product page. Include FAQ schema on category pages. The more structured data AI agents can parse, the more accurately they can recommend your products.

5. Reduce friction in your own funnel

If Perplexity's success proves anything, it is that the winners in AI commerce reduce friction rather than add it. OpenAI added friction by building a new checkout. Perplexity reduced friction by integrating PayPal. Apply the same principle to your store.

What to do: Audit your own checkout friction. Enable accelerated checkout options (Shop Pay, PayPal Express, Apple Pay). Minimize form fields. Ensure mobile checkout is fast and clean. When AI agents send traffic to your store, the last thing you want is a clunky checkout that wastes the referral.

The bottom line

ChatGPT's checkout failure is not a cautionary tale about AI shopping. It is a cautionary tale about ignoring commerce fundamentals. The AI can be world-class. But if product data is stale, if the checkout strips away merchant trust, and if the experience adds friction instead of removing it, customers will abandon.

The merchants who benefit from the next phase of AI commerce -- where AI platforms drive discovery and merchants own conversion -- will be the ones who have already done three things:

1. Cleaned their product data so AI agents represent them accurately
2. Measured AI as a channel so they know what is working
3. Optimized their own checkout so AI-referred traffic converts

These are not future tasks. They are today's priorities. The agentic commerce wave is not slowing down because one checkout experiment failed. It is accelerating toward a model where data quality, merchant trust, and checkout simplicity win.

Run an AI ranking report to see where your brand appears in AI shopping results today. Then audit your structured data, fix your tracking, and own the checkout experience that turns AI discovery into revenue.