The AI Shopping Channel Mix: Google, Marketplaces, and ChatGPT

For the last 15 years, the default ecommerce strategy deck has looked the same:

• Win Google (SEO + Shopping + Performance Max)
• Win Amazon (Marketplace + Sponsored Products)
• Layer on paid social and email/SMS to retarget

The assumption was simple: most product discovery starts in a search bar (usually Google) or a marketplace (usually Amazon), and your job as a brand is to be there, bid there, and rank there.

That assumption is now under pressure.

Generative AI and conversational agents – especially ChatGPT – are quietly becoming a third entry point into the shopping journey. OpenAI has launched Instant Checkout in ChatGPT, starting with Etsy and soon Shopify merchants, so U.S. users can ask what to buy and check out without leaving the conversation.

This post is about how to think about that in the context of your existing channel mix.

• What role does Google still play?
• How much power have marketplaces like Amazon consolidated?
• Where does ChatGPT – and AI agents more broadly – fit in?
• And what should a sane ecommerce leader actually do in 2025–2026?

1. The current landscape: Google vs Marketplaces

Google: still the front door for many shoppers

Despite the narrative that “everyone starts on Amazon,” search engines still matter:

• In a Klarna study, 44% of shoppers said they started their shopping journey on a search engine, with 41% going directly to online stores like Amazon or a brand site.
• Google remains the default place to research brands, read editorial content, and compare categories where Amazon’s assortment is weaker (e.g., niche DTC, certain B2B products).

Google’s Shopping ads and free listings still drive massive intent. But the nature of that traffic is changing:

• Google is embedding AI Overviews and other generative features directly into search results, especially for complex queries.
• For younger demographics, especially Gen Z, Google’s mix of AI search and tools like Lens is regaining share from Amazon: in one Morgan Stanley survey, the share of Gen Z who used Google first when they already knew what they wanted to buy rose from 21% to 30% in six months, while Amazon’s fell from 41% to 34%.

In other words: Google is not “dead.” It’s mutating into its own form of AI-driven shopping assistant. But it’s still fundamentally a search surface where merchants win or lose based on:

• structured data (schema.org, Merchant Center feeds),
• paid bids, and
• classic ranking signals.

Marketplaces: Amazon as a default starting point

Multiple independent reports support the same core fact: more than half of US consumers start product searches on Amazon:

• Jungle Scout’s 2024 Consumer Trends report found that 56% of consumers start their product searches on Amazon, compared to 42% on search engines.
• Other analyses put the figure in the 50–63% range, with Amazon increasingly described as “the default ecommerce store for the majority of purchases.”

This is not just about search. Amazon has:

• customer accounts and stored payment methods
• Prime shipping expectations
• reviews and Q&A as embedded trust signals
• sophisticated advertising infrastructure

As a result, Amazon is often the first and last touch point: discovery, comparison, and purchase all happen inside the marketplace.

And Amazon is not alone: Walmart, Temu, Shein, niche marketplaces, and vertical aggregators all play similar roles in their respective categories.

So pre-ChatGPT, your channel mix looked like:

• Google/search: generic discovery, category research, and some purchase journeys
• Amazon/marketplaces: heavy share of product-first searches and conversion
• Direct-to-consumer: loyal repeat buyers, brand-led experiences, and higher-margin sales

Now we have a new player.

2. Enter ChatGPT: the third entry point

In late 2025, OpenAI began rolling out Instant Checkout in ChatGPT, powered by the open-source Agentic Commerce Protocol (ACP).

A few key facts:

• U.S. ChatGPT users can now buy directly from Etsy sellers in-chat, across free, Plus, and Pro accounts.
• Support for Shopify merchants is “coming soon,” meaning over 1 million Shopify stores will eventually be able to sell through ChatGPT conversations.
• ChatGPT’s product recommendations are ranked based on price, availability, quality, seller prominence, and checkout ease – not on paid placement.

On the payments side:

• Stripe implemented the Delegated Payment Spec to let ChatGPT send a one-time payment request to PSPs without seeing raw card data.
• PayPal has announced it will adopt ACP so millions of users can check out via PayPal directly in ChatGPT.

This means ChatGPT is no longer just an “advice” surface. It is:

• a search interface (natural language prompts instead of keywords),
• a comparison engine (it can read product feeds, reviews, specs), and
• a checkout surface (Instant Checkout embedded in the chat).

How is this different from Google or Amazon?

• It is intent-rich: the prompt often encodes budget, constraints, use case, urgency, and preferences.
• It is cross-merchant by design: the agent can consider products from multiple merchants, then route an order to the one that best fits.
• It is conversation-first: follow-ups (“I have wide feet”, “make it vegan”, “prefer black”) refine the set in ways that are painful in traditional interfaces.

From a channel perspective, ChatGPT is like a blend of:

• Google’s role in discovery, and
• Amazon’s role in conversion,

but mediated by a model that chooses for the user.

3. How to think about the mix: complementary, not zero-sum (yet)

It’s tempting to think of this as a simple “Google vs Amazon vs ChatGPT” fight. In reality, they currently play different roles:

• Google is where people search when they want breadth and are willing to click around.
• Amazon/marketplaces is where they go when they want depth in a known context (Prime, reviews, familiar filters).
• ChatGPT is where they go when they want done-for-them decision-making in natural language.

A single shopper might:

1. Use ChatGPT to narrow the field (“What’s a reputable brand for beginner espresso machines?”)
2. Check Google to read independent reviews and guides.
3. End up buying on Amazon or a brand site depending on price, shipping, or loyalty.

You shouldn’t think “we’re dropping Google to go all-in on ChatGPT.” Instead, treat ChatGPT as:

A new acquisition and conversion surface that will sit alongside search and marketplaces – and borrow mechanics from both.

The job of your team is to:

• understand what queries and use cases are likely to hit ChatGPT first
• ensure your products are eligible to be considered and sold there
• attribute and compare performance against Google and Amazon over time

4. Where ChatGPT is strong (and where it’s weak)

Strong: complex, conversational queries

ChatGPT shines when the user has:

• multiple constraints (“vegan, under $80, ships to Germany, good for sensitive skin”)
• fuzzy intent (“something nice for a coworker who likes coffee”)
• desire for explanation (“why this vs that?”)

These are exactly the queries where traditional search results are overwhelming or noisy.

Strong: “one good option” decisions

Because ChatGPT effectively pre-filters, you may only get one or three spots in its response. Getting one of those is incredibly valuable, especially in higher-ticket or more considered purchases.

Weak (for now): real-time bargain hunting

If the user’s primary objective is “find the absolute cheapest price,” marketplace and comparison engines still have an edge:

• They are better wired into pricing feeds across dozens of merchants.
• They focus heavily on sort-by-price and dense grid comparisons.

Over time, this may change as product feeds and ACP integrations become richer – but in the near term, expect ChatGPT to dominate “help me choose” more than “show me the absolute lowest price.”

5. Practical implications for your channel strategy

A. You still need Google and marketplaces

Nothing about Agentic Commerce changes the fact that:

• Google will continue to be a starting point for many journeys.
• Marketplaces like Amazon will likely remain the default for a huge class of search-intent shoppers and categories.

You’re not unplugging anything.

B. You do need an AI/agent line item in your planning

What should that look like?

At minimum:

1. Data readiness

   • Audit your product feeds (to Google, Amazon, others).
   • Compare them to OpenAI’s Product Feed Spec.
   • Identify missing required fields, inconsistent attributes, stale inventory, and broken links.

2. Eligibility readiness

   • Decide which SKUs should be eligible for ChatGPT commerce first (high-margin, low-return, simple to ship).
   • Ensure those products have complete, normalized data in your feed.
   • Plan for 15-minute updates to keep price and stock in sync.

3. Checkout readiness

   • Read through OpenAI’s Agentic Checkout Spec.
   • Confirm your systems can handle:
     • creating/updating cart sessions via API
     • calculating tax and shipping on the fly
     • accepting delegated payment tokens (via Stripe, PayPal, or your PSP)
     • emitting order lifecycle webhooks (created, updated, refunded, etc.)

4. Measurement readiness

   • Tag orders originating from ChatGPT/agents with a source or channel field.
   • Track GMV, AOV, and return rate for this source.
   • Compare performance to:
     • Google organic
     • PLA / Shopping ads
     • Amazon marketplace sales

This doesn’t mean shifting huge budget immediately. It means putting down the plumbing and instrumentation so that when the traffic shows up, you can see it, serve it, and learn from it.

6. How Pesto can help (briefly)

You can absolutely build the Agentic Commerce layer yourself – many large retailers will.

But for most mid-market and large ecommerce teams, the hard part isn’t just reading the spec once. It’s:

• keeping up as OpenAI evolves the Product Feed and Checkout specs
• ensuring your data stays clean, normalized, and compliant across thousands of SKUs
• hosting and monitoring feeds so you know which products are eligible, auto-fixed, or being dropped
• coordinating with PSPs to ensure delegated payments succeed and errors are handled gracefully

Pesto exists to sit in that middle layer between:

Your Store ↔ Pesto ↔ ChatGPT

We:

• ingest and normalize your catalog (Shopify alternatives included)
• map it to OpenAI’s Product Feed and Checkout specs
• host a clean, always-on feed
• keep it in sync every 15 minutes
• surface which SKUs are eligible, auto-fixed, or quarantined

Whether you use Pesto or build your own infra, the strategic point is the same:

Google and marketplaces aren’t going away – but AI agents like ChatGPT are becoming a third, high-intent entry point to your brand.

You don’t want to be the last one in your category to show up there with a working feed and a working checkout.

The brands that treat ChatGPT as part of their channel mix – not a toy – will be the ones that own the next wave of AI-driven revenue.