ChatGPT Can Now Buy Things. What Marketers Should Do Next.

OpenAI added Instant Checkout to ChatGPT. U.S. users can purchase from Etsy sellers directly in the chat, with Shopify merchants “coming soon.” (OpenAI)

This is a practical shift. Last year, 700 million conversations occurred between ChatGPT and a user considering a purchase (OpenAI). Although ChatGPT could influence or recommend the purchase, it could not help you make it. The user had to visit the seller’s website and checkout there. 

With Instant Checkout, ChatGPT cuts the seller’s website out of the equation. When the conversation and checkout live in ChatGPT, buyers need not browse your site before acting. When buyers skip a seller’s website altogether, the sellers cannot offer deals, promote new products, or advertise a loyalty program.

ChatGPT (and other AI models) are evolving into commercial middlemen. They will stand between brands and their prospective customers. Brands will rely on ChatGPT to sell their products rather than persuading customers directly with their website messaging.

Today, this change only affects E-Commerce. However, a similar change is coming for B2B brands.

Below is a concise, marketing-oriented checklist to adapt your funnel to this change.

1) Assume “book a demo in chat” is next—and prepare content for it

While the current feature is retail-focused, the underlying Agentic Commerce Protocol passes order details to the merchant’s systems. The same pattern could support actions like request a quote, start a trial, or book a demo without a site visit. Treat this as a near-term planning scenario. (OpenAI Developers)

What to do

• Write a two-sentence, model-legible pitch that survives paraphrase: who you’re best for, in what situations, with what outcome.

• Publish a decision path the model can reuse: “Choose us when ___; choose X when ___.” Keep it explicit and source-backed.

• Create a friction-free handoff: calendar URLs, trial links, starter plans—anything a model can present as the next step.

2) Make claims verifiable 

Models weight evidence. ChatGPT ranks merchant options on factors like availability, price, quality, and support for Instant Checkout, and it draws on structured sources for results. For B2B, expect a similar emphasis on verifiable signals over slogans. (AP News)

What to do

• Tie differentiators to citations: docs, certifications, benchmarks, case studies with concrete numbers.

• Expose structured data: specs, integration lists, pricing tiers, SLAs, compliance claims. Use clean pages and feeds the model can ingest. (OpenAI Developers)

3) Optimize for recall, not just readability

Models compress. Long positioning statements get reduced to a sentence or two when recalled in conversation.

What to do

• Canonicalize three short proofs you want retrieved (e.g., “fastest deployment in category—median 2.5 days,” “SOC 2 Type II + HIPAA,” “top-quartile retention in audited cohort”).

• Standardize the phrasing across your site and materials to increase the chance the model repeats it consistently.

4) Measure beyond site traffic

If a buyer completes actions in-chat, site sessions and pageviews will undercount demand capture.

What to do

• Track mentions → recommendations → actions (e.g., whether ChatGPT surfaces you unprompted, recommends you against peers, and completes a trial/demo route).

• Instrument post-chat paths (trial activations, welcome emails, CRM stage changes) so you can infer in-chat conversions even when the first touch isn’t a pageview.

5) Start thinking about “AI Relations”

There is a lot of chatter about “AEO” and “GEO” as the solution to this problem. This suggests that getting recommended by AI models is like SEO, which is comforting to marketers who are familiar with SEO. 

However, this abstraction misses the mark. AI models are less like search engines than key opinion leaders for your customers. Models make recommendations, have opinions, and every query/prompt ends up with a different result.

Instead, use analytics to learn how AI models talk about your brand, then ship new content and proofs to close them. (I wrote a separate piece on this: AI Relations vs. AEO/GEO).

Bottom line

Buying inside ChatGPT moves a meaningful slice of persuasion and conversion into the conversation itself. Treat the model like an influencer that you have to brief with concise, source-backed narratives—then make it easy for the agent to complete the next step, even if a prospect never hits your homepage.

Further reading on the release:

• AP News

• Investors