Disclaimer:

In the context of this post, I’m referring to shopping and checkout experiences that occur natively in Answer Engines (AE) like ChatGPT. These experiences aren’t really “agentic” (yet) but the term persists. The term “Agentic Commerce” is in the zeitgeist. Folks often use it to mean commerce processes with agents involved (despite their mainly being assistive, rather than agentic).

Answer Engines Are Monetizing Transactions

OpenAI has confirmed to Forrester that the answer engine is now charging a 4% fee to Shopify merchants when customers complete checkout within the answer engine. Shopify’s partnership with OpenAI enables merchants to automatically send a feed of their catalog data to OpenAI. (This is starkly different from other models in which answer engines crawl merchants’ websites, scrape product data, and list the products in the answer engine.) Of course, other commerce solutions and Product Information Management (PIM) solutions have built feeds as well. But Shopify is also participating in ChatGPT’s  “Instant Checkout,” which lets customers complete their order within ChatGPT.

This fee isn’t surprising: I’ve been waiting to see how Answer Engines would monetize activities and transactions. However, Shopify already charges its own fees to merchants for the transactions that use Shopify Payments – commonly 2.9% plus $0.30 (though it can vary and is on par with universal fees for credit card payments). Shopify Payments is required for native checkout in ChatGPT.

Total Costs of Selling In Answer Engines Are Not Yet Clear

We already know that merchants will pay 7%+ per transaction for these “agentic commerce” transactions. This doesn’t include additional potential tech and other costs of the transaction.

Merchants have many unanswered questions around potential costs to come:

  1. Will ChatGPT’s fees apply only to Shopify merchants or to all merchants enabling the native checkout feature? Will the fees be the same for all merchant systems?
  2. What would Google Gemini, Perplexity, or Microsoft Copilot charge for similar experiences? Will they use these fees as a point of differentiation or an opportunity for monetization?
  3. What will be the additional costs for managing catalogs specifically for this channel? (E.g., I suspect Shopify may charge additional fees to customers who don’t use the commerce engine but instead create catalogs in Shopify to send to Answer Engines, although no fee is announced yet.)
  4. Will other tech like order management systems (OMS) jump into the mix to manage inventory levels for answer engines? How will these vendors charge for this service?

Answer Engines Are A New Digital Channel That Need Strategy, Not Panic

Pump the brakes a bit – you need strategy first. This moment of almost-Agentic Commerce harkens back to the early days of drop shipping (both strategies and technology). Brands and merchants felt urgency then, too, to figure out how to get their products into this new channel and often learned the hard way which channels were actually profitable.

Remember: we’ve been syndicating to additional digital channels for decades. We just need to find what’s unique about this channel, learn which customers it’s best suited for, and adjust accordingly.

My colleagues and I have lots more to come on Agentic Commerce (such as it is today) and the frameworks and principles that should guide your strategies for it. But strategy is paramount.

Wanting to “do Agentic Commerce” isn’t a strategy. Craft your strategy first, including:

  1. The P&L: What’s the total cost of selling in each channel? The risk of margin erosion is real. Factor it into product prices or avoid selling products with minimal mark-up in answer engines. The approximate 7% is still about half of Amazon’s cut, but marketplaces requires a completely different strategy. They also might attract a completely different shopper profile. We don’t know what other fees may come, including from systems that send catalog feeds to the Answer Engines. Add AE channels strategically, and monitor the P&L per channel to identify unexpected margin erosion early.
  2. Customer fit: Are your target customers in the channel? Shoppers haven’t widely adopted the use of answer engines for shopping. According to Forrester’s December 2025 Consumer Pulse Survey, 36% of US online adults aged 18-29 have used ChatGPT in the past month to search for products they’re interested in buying. That slips to 31% for those aged 30-49, and falls off to a measly 11% for those 50 and older.
  3. Product control: How do you choose which products to sell in each channel? Many answer engines are currently “crawling” commerce websites, and “scraping” product data to include in synthesized answers. That means that you as a merchant have limited control at best over the content (including product information, images, pricing, and availability) that answer engines synthesize for shoppers. Learn from your drop shipping experience: Merchants that drop ship strategically manage assortments by channel, choosing which products to offer and where. They maintain separate product descriptions and assets to ensure they’re presenting those appropriately for each channel. The “pull” of crawling and scraping removes some control, though Answer Engine Optimization (AEO) presents some opportunities. Tech and services vendors will emerge to offer more controlled methods for “pushing” products strategically into answer engines. (And, remember – it’s okay not to be first.)
  1. Third party support: Which providers will streamline AE selling for you? You don’t have to build architecture from scratch. Shopify is making its catalog available to merchants who don’t use its commerce engine. Commerce (FKA BigCommerce) is leveraging its Feedonomics acquisition to enable similar strategies. Merchants can pull in their commerce catalogs, adjust them as they’d like, and push them into Perplexity for now (and soon, to Google Surfaces), whether or not they use BigCommerce as their commerce engine. Commerce services providers also are developing catalog feeds (for a fee, of course). Watch closely as providers in the market offer tools that will support your foray into AE selling.

For my tech vendors: move (very) fast, watch the market, and don’t act in panic. While I advocate calm for brands and retailers, I know that for vendors, this moment is a race, for sure, and it’s clearly visible in the market. Vendors: don’t launch poor experiences prematurely just to get something to market. Instead, find early relationships and monitor the market for (currently) exclusive partnerships to expire and open new opportunities.

Spoiler for future reports/posts: One of the things that’s remarkable is the fact that the commerce solution is the hub for this data, particularly when it’s scraped. Product Information Systems (PIM) that typically syndicate product data to selling channels are building feeds to answer engines. But crawling and scraping websites to obtain product data bypasses some data management systems. This will play out in new ways.

If you’d like to understand how all of this impacts you and what to do about it, set up a call with me to talk it through! And if you’re interested in specific areas outside my coverage, please see below for some of my expert colleagues and their respective coverage.

  • Emily Pfeiffer – Agentic Commerce in general, Commerce Solutions, Order Management Systems (OMS), Commerce Search & Product Discovery (site search), drop shipping.
  • Chuck Gahun – Commerce Services, Content Management Systems (CMS), Agentic Content.
  • Lily Varon – Payments, Billing, Subscriptions, POS, Agentic Payments.
  • Nikhil Lai – Answer Agent Optimization, Search Engine Optimization, Zero-Click Search impacts.