Can merchants really sell inside ChatGPT today? Busting 5 myths about agentic AI in commerce

The headlines about “buying inside ChatGPT” and other AI platforms are loud. The reality is much quieter.  

Our AI experts unpack the five common myths that are causing confusion, detail where the tech actually stands today, and share the must-ask questions for any retailer or brand exploring where to go next with agentic AI integrations. 

Myth 1: Merchants are selling directly in LLM platforms (such as ChatGPT, Perplexity, Gemini, and Copilot today) from search to checkout.  

Truth: As of now, there is no checkout-in-chat option that everyone can turn on in an LLM. The closest anyone has come is the Instant Checkout option in ChatGPT, which is only available in the U.S., and only for single-item purchases of Etsy products that include a “Buy now” option.  

The Etsy and ChatGPT Instant Checkout flow was built on an Agentic Commerce Protocol co-developed with Stripe. Here’s how it works: 

1. A U.S. shopper finds an eligible Etsy item in ChatGPT and taps Buy (two caveats: this is available for single-item purchases only, and only for sellers who choose to opt-in. This means not every Etsy suggestion offered by ChatGPT is purchasable within ChatGPT.) 
2. ChatGPT collects the order details, shipping, and payment confirmations inside the conversation.  
3. Without sending the shopper out of the conversation, ChatGPT passes the order plus a scoped payment token to Etsy.
4. Etsy authorizes the payment with the sellers’ existing processor and the shopper gets an in-conversation confirmation.  
5. Merchants pay a small fee on completed orders. Any follow-up (shipping updates, returns) comes through their normal channels.   

Etsy is the first wide-scale pilot, and Shopify says a staged rollout will continue with Glossier, Skims, Spanx, and Vuori; Walmart, the world’s largest retailer, has also been named as an upcoming launch partner. In parallel, Salesforce announced an expanded OpenAI partnership that brings Agentforce/Commerce Cloud into ChatGPT via Apps in ChatGPT and ACP-powered Instant Checkout—effectively enabling Salesforce Commerce Cloud merchants to integrate into ChatGPT directly. 

In addition to ChatGPT’s plans, other agentic AI search-to-checkout is being developed as well: Perplexity’s Buy with Pro lets U.S. Perplexity Pro users purchase a limited set of items with saved payment and shipping. But inventory coverage is thin, and most queries still kick back to the merchant’s website for checkout. 

These announcements made big headlines. And they are exciting, because they validate agentic AI’s model and momentum. But they are controlled pilots, not blanket access, and in some cases not yet live. They’re still U.S.-first and far from universal coverage across categories, price points, and complex use cases. 

That’s the current state of agentic AI search-to-checkout commerce: limited but growing quickly. In many cases, the headlines are focused on roadmaps. At Rithum, we’re making sure our clients are ready now, so they can meet the soon-to-come expansions without any hiccups. 

Myth 2: LLMs are the next sales channel right now. 

Truth: Today, agentic AI is used by shoppers for discovery, recommendations, and engagement. But (except in a few niche cases) when the shopper shifts to buyer, they still must go to your site or a marketplace.   

Even most owned LLMs route shoppers back into a standard web checkout. And the few non-owned LLMs that include checkout offer rare, immature end-to-end examples with thin assortments (see the first myth above for those details). 

Much of the “LLMs as a sales channel” excitement points to future rollouts. But it’s true that new partnerships, API pilots, and early checkout integrations are setting the stage for an inevitable shift. 

For now, treat LLMs as search and discovery platforms, but prepare for its fast-approaching sales channel-optimization future. For both realities, the best route is the same: connected catalogs with optimized visibility for AI-driven search results.  

Myth 3: If you’re not in a pilot now, it’s too late. 

Truth: Agentic AI check-out is far from mainstream. Right now, your focus should be on discoverability by LLM search, so you can be ready when LLM shopping expands.  

LLMs don’t fix messy catalogs or operational gaps: they expose them. The quality of what shoppers see (titles, images, variants, pricing, availability, reviews) still depends on the quality of your product listings. And even in LLMs, when a shopper clicks “purchase,” the same old pipes—for payments, taxes, fraud prevention, fulfillment, returns, support, etc.—will still have to run smoothly.   

When end-to-end shopping in LLMs is truly ready, here’s the minimum scaffolding that will be needed to make it work: 

• Fraud integration and payments: To process payments with Stripe and ChatGPT, use Agentic Commerce Protocol to make the process straightforward. Expect the usual realities: auth declines, fraud checks, 3D Secure/SCA where required, and disputes. An agentic AI front end doesn’t erase back-end frictions and you being the merchant of record. 
• Catalog and content health: In LLM results you can’t pay-to-place (yet, at least), so clean, complete, and consistent feeds are your best ranking strategy. Start now by maintaining accurate, structured product data (pricing, inventory, variants, shipping promises) and focusing on high-quality, real consumer reviews.  
• Operations and service: You’re still the merchant of record. Fulfillment speed, order changes, cancellations, returns, warranties, and support SLAs all remain yours. If those break, so does your customer experience. 
• Attribution and analytics: In-chat discovery and checkout are a whole new world for ROI measurement. Plan for event tagging, order source stitching, and deciding now how you will count these new types of conversions. 
• Policy and edge cases: Subscriptions, preorders, regulated or age-gated goods, complex bundles, and cross-border tax/compliance that need special handling will likely not be supported for some time. 

When there is a wider roll-out of agentic AI check-outs, these will all be unowned experiences you can’t control. Your shoppers’ experience with your brand will hinge on accurate, structured, fresh product feeds with price, availability, and variants, and your data’s ability to respond to the LLM’s calls for inventory, shipping, tax, and order status. In short: clean data feeds. Clean data feeds will also will make you more discoverable to AI in general, which is where you’ll get the most ROI now (and in the future). The best preparation you can do now is to make your catalog agentic-ready.  

Myth 4: Generative Engine Optimization (GEO) will replace SEO soon. 

Truth: GEO is a shift, not a flip. Classic SEO—clean architecture, relevant content, smart internal linking—still matters. And the good news is that most of what you built for SEO creates a good framework for GEO. The major difference is that SEO was about optimizing content for search engines (using repeated keywords, for example), while GEO is about optimizing users’ specific use cases.  

This means you need better quality data that aligns with a customer’s needs. Structured product data, trustworthy specs, consistent price/availability, and verified reviews that answer real shopper questions are more valuable than ever. 

While GEO is gaining speed, some large retailers are throttling or fully blocking the AI crawlers that GEO relies on. That hurts you on three fronts:  

• AI visibility: When a retailer slows or block AI bots, they can’t crawl enough of your catalog to get specific details (such as brand/model/variant) or refresh listings, so you’re excluded from more answers.  
• AI accuracy: The costs for bots to scrape data is significant. For them to scrape on hourly basis would be cost prohibitive, which means they are not actively looking for updates. If your product data changes, you’re at their mercy for when they might get the update. 
• Product eligibility: Rich placements (configurable cards, in-chat cart/checkout, live delivery estimates) require verified, current fields. Blocking AI bots starves engines of the confirmations they need, so you fail their trust checks and get downgraded to plain-text mentions. 

Even when crawlers are allowed, LLMs struggle with variants, fast-moving prices, and inventory freshness. This is all to say, simply assuming your pages will be accurately scraped on a retailer’s page won’t cut it in the GEO-world. Structured feeds, order/stock webhooks, and commerce protocols that deliver authoritative, up-to-the-minute data so agents can answer accurately are your best bet. 

Don’t abandon SEO, and don’t ignore GEO. Get used to living this messy middle—earning placement in traditional results while feeding agentic AI engines precise, trustworthy data. Do this well and you’ll get better clicks from classic search and steadily compounding visibility in agentic AI results.  

Myth 5: It’s impossible to tell what’s hype and what’s reality when vendors pitch us on their agentic AI capabilities. 

Truth: Hopefully the level-setting above has helped, but it all comes down to asking the right questions. Start with these:  

1. Is it live for all users or invite-only? 
2. Does the LLM list the integration publicly? 
3. Is payment completed inside the LLM or redirected to your site? 
4. How do we pass live inventory, pricing, and product content updates? 
5. How do we optimize our placement within the LLM platform? 
6. What reporting will we get, including prompts, impressions, and conversions? 

The biggest truth: Rithum can help. 

Rithum offers a future-proof way to participate in the LLM platform shift—starting with clean data pipelines today and preparing for agent-driven transactions tomorrow.  

Rithum enables you to be found by agentic AI programs by directly feeding LLMs with structured, real-time product data. This avoids outdated or incomplete information appearing in AI-assisted shopping conversations. 

Here’s how it works: 

• Product feed compilation: Rithum aggregates and structures your product catalog. 
• LLM optimization: Content is enhanced to meet the requirements of LLMs, such as ChatGPT and Perplexity. 
• Feed delivery: Rithum provides compliant feeds directly to these AI platforms. 

This approach ensures that AI assistants present accurate pricing, availability, and product descriptions, and that any shoppers working within AI who find your product have a smoother pathway to purchase. 

Our agentic AI roadmap includes deeper support to continuously meet the commerce industry as it moves more into full AI experiences, including direct LLM integrations for more control and real-time updates, GEO optimization to improve ranking and visibility, and LLM monitoring and reporting to track mentions, conversions, and competitive positioning.  

You are not too late to agentic AI. The agentic commerce work of today lives in the quiet, unglamorous data feeds. Getting the data pipes ready now ensures you’ll have clean listings later. Talk to us about how it works, and how we can help. 

Sources cited: 

• Stripe powers Instant Checkout in ChatGPT and releases Agentic Commerce Protocol codeveloped with OpenAI  
• OpenAI partners with Walmart to let users buy products in ChatGPT, furthering chatbot shopping push) 
• Shop like a Pro: Perplexity’s new AI-powered shopping assistant 
• I tried shopping for a gift using a new ChatGPT feature. It only got more confusing as I searched. 
• Amazon quietly blocks AI bots from Meta, Google, Huawei and more 
• The explosive rise of generative AI referral traffic