Everyone Says AI Agents Will Do the Holiday Shopping This Year. What Do I Change Before Q4?

If you sell online, you have watched two things happen at once this year: fewer humans clicking through from search, and a strange new trickle of visitors who arrive already knowing exactly what they want. Those visitors were briefed by an AI assistant before they ever touched your site. The holiday season is the first Q4 where that traffic is big enough to matter, and August is when you still have time to prepare for it calmly instead of in a November panic.

My traffic is down but sales are holding. Are AI assistants already shopping my store?

Quite possibly yes: Adobe Analytics measured AI referred traffic to US retail sites up 393 percent year over year in the first quarter of 2026, and by March that traffic converted 42 percent better than direct traffic. The pattern sellers describe, fewer visits but steady orders, is exactly what this shift looks like from the inside. An assistant does the browsing, comparing, and shortlisting on the buyer's behalf, then sends over one highly qualified human who mostly just confirms the purchase. Adobe's data shows those visitors also engage 12 percent more and spend 48 percent longer on site once they land, with revenue per visit running 37 percent higher than other traffic. The holiday preview already happened: during November and December 2025, AI referred retail traffic was up 693 percent year over year. That was the rehearsal. This Q4 is the show.

ChatGPT shut down its buy button in March. Do I still need to care about any of this?

Yes, because the buy button was never the part that decided who gets the sale: the recommendation was. OpenAI launched Instant Checkout purchasing inside ChatGPT in February 2026 and quietly removed it on March 4, with only around 30 merchants live, after the experience suffered from inaccurate pricing and inventory data pulled from web scraping, as reported by Digital Commerce 360. The company has since repositioned around product discovery and merchant apps, and the Agentic Commerce Protocol it codeveloped with Stripe lives on underneath. Read that failure carefully, because it is the most useful free lesson a seller will get this year. The checkout plumbing failed on bad product data. The assistants are still answering millions of what should I buy questions every day, and they answer them by reading whatever your listings and feeds actually say. If your data is wrong or thin, you are not penalized with a lower ranking. You are simply never mentioned.

Want the complete system for this? The full playbook for making your catalog the one the machines quote is Sell to the Machines: The AI Shopping Agent Playbook.

What do I actually fix before Q4 so AI assistants recommend my products?

Fix your product data first, because every documented agentic commerce failure so far traces back to stale prices, missing specs, or pages a machine could not parse. If that is roughly what you typed into a chatbot, here is the checklist you wanted:

• Make every listing spec-complete in text. Dimensions, materials, compatibility, quantity per pack, care instructions. An assistant cannot read the beautiful infographic where you buried the size chart. If a spec only exists in an image, to the machine it does not exist.
• Ship valid Product structured data. On your own site that means Product JSON-LD with name, image, price, and availability at minimum. This is the same markup Google merchant listings already require, so you may be closer than you think.
• Write plain, factual copy that answers buyer questions. Assistants quote listings that state facts. Will this fit, what is it made of, what is in the box, how long does shipping take. Adjectives do not get retrieved. Answers do.
• Sync your feeds everywhere your catalog lives. The Instant Checkout shutdown was triggered by price and inventory mismatches. If your marketplace listing, your site, and your feed disagree, the assistant either picks the wrong number or skips you.
• Put shipping and return policies where a machine can read them. Assistants increasingly factor delivery time and return terms into recommendations, especially for holiday deadlines.

One caveat before any of this helps: an AI assistant cannot recommend a listing that a marketplace has suppressed. If your products keep getting flagged or removed before a machine ever sees them, that is a different fight, and The Restricted-Category Seller Playbook covers the appeal templates and terminology fixes for it.

How do I know if any of this is working before the holiday rush?

Segment AI referrals in your analytics now and judge them on conversion rate and revenue per visit, not raw volume. Referrers from the major assistants show up in most analytics tools and the volume will look small, which fools sellers into ignoring it. The number that matters is what that traffic does after it arrives. Adobe's March 2026 figures, 42 percent better conversion than direct, are an average across US retail. Your own segment tells you whether your listings are being represented accurately. If AI referred visitors bounce, the assistant promised them something your page does not clearly deliver, and that is a fixable copy problem, not a traffic problem. Check it monthly between now and October, and you will walk into Q4 knowing exactly how the machines describe your store.

Want the complete system?

This is the checklist. The complete method, how each major assistant retrieves and ranks products, the listing rewrite framework with before and after examples, structured data templates you can paste, marketplace-by-marketplace playbooks for Amazon, Shopify, eBay, and Etsy, and the monthly monitoring routine, is in Sell to the Machines: The AI Shopping Agent Playbook for Amazon, Shopify, and Marketplace Sellers. Written for working sellers, not developers. Instant download.

→ Get Sell to the Machines ($79)

Frequently asked questions

Do I need to add special code to my store for AI shopping agents?
Mostly no. The highest-impact work is complete product information in plain text and standard Product structured data, which mainstream Shopify themes and marketplace listings already support. Protocols like the Agentic Commerce Protocol matter later; data quality matters now.

I only sell on Amazon, not my own site. Does any of this apply to me?
Yes. Assistants read marketplace listings too, and Amazon's own AI shopping features rely on the same listing content. Spec-complete bullets, accurate attributes, and consistent pricing help you inside and outside the marketplace at the same time.

My product photos carry most of my listing. Is that a problem for AI assistants?
For the machine reading your page, yes. Keep the photos for humans, but duplicate every fact they contain, sizes, materials, contents, compatibility, as plain text on the listing so an assistant can retrieve and quote it.

Whatever happened to buying things directly inside ChatGPT?
OpenAI removed its Instant Checkout feature on March 4, 2026 after pricing and inventory accuracy problems, with about 30 merchants live, per Digital Commerce 360. Discovery inside assistants continues to grow, and the takeaway for sellers is that accurate product data is what the next version of in-chat buying will be built on.