Shopify agentic commerce: what to fix before Black Friday

Key takeaways

  • Adobe expects AI-referred holiday traffic to US retail sites to grow about 130%, from a small base. For an example $10M brand, that is about $7,000 to $10,400 of BFCM revenue.
  • The larger exposure is what agents tell shoppers about your shipping cutoffs and returns before they reach your product page.
  • Before Thanksgiving on November 26, fill in the Agentic knowledge base and move your quiz answers into product data.
  • Add an AI Assistants channel in GA4 above Referral, and read conversion and average order after Cyber Monday before changing the plan.
  • Release dates and which agents support checkout inside the chat are still unsettled, so check the Agentic sales channel in your own admin.

Shopify agentic commerce lets AI assistants such as ChatGPT, Copilot, Google AI Mode and Gemini show your products and, in some cases, check shoppers out in the chat. Before Black Friday, fill in Shopify's Agentic knowledge base with holiday shipping cutoffs, return window and gift policy, complete your product data, and track AI referrals as their own GA4 channel.

Open ChatGPT and ask whether your best seller will arrive by December 24 if someone orders it on December 18. Then ask Gemini the same question.

If either one guesses or quotes last year's cutoff, some of your Black Friday shoppers will get that answer before they reach your product page.

Shopify's Fall '26 Edition pushes checkout into AI assistants and Google AI Mode. Adobe expects traffic from AI chat and browser referrals to US retail sites to grow about 130% this holiday season, with the biggest jump on Thanksgiving (+159%).

That growth starts from a small base. For most Shopify brands, AI referrals will be a small share of BFCM sessions. The exposure is in the answers. A shopper can read what an agent says about your shipping and returns, decide, and never click through, so the traffic number understates how many shoppers the agents reach.

The fix takes about an hour:

  1. Fill in Shopify's Agentic knowledge base with your holiday shipping cutoffs, return window and gift policy.
  2. Add an AI referral channel in GA4 so you can see whether that traffic converts for your brand.
  3. Ask the main assistants about your top products and fix what they get wrong.

Start with the knowledge base. Coverage of the Fall '26 Edition places it under Sales channels > Agentic in Shopify admin, where you approve what agents learn about return windows and shipping. Thanksgiving is November 26, so do it this week.

Know what Shopify changed this fall and what is still unclear

According to Shopify's Fall '26 Edition announcement and recaps of it, three things changed:

  • Shop Pay checkout works inside third-party AI assistants.
  • Merchants can sell through Google AI Mode and the Gemini app with Shopify-powered checkout.
  • A knowledge base under the Agentic sales channel lets you approve what agents say about returns and shipping.

Three details are not settled.

Release date. One recap says the edition shipped October 28 with 150+ updates. Another says the checkout tools went live September 28.

Which agents check out inside the chat. Shopify's own help pages have described ChatGPT as a referral channel, where the shopper clicks through to your store to buy. Other coverage describes direct checkout settings for Copilot, Google AI Mode and Gemini. Check the Agentic sales channel in your admin for what is on in your store.

The conversion edge. Adobe reports that AI-referred visitors convert better. Write-ups put the edge at 48% in some places and 40% in others.

Work out what AI referrals are worth to your store

The Adobe 2026 holiday forecast expects $275.1B in US online spend from November 1 to December 31. The 130% applies only to AI referrals, a small slice of that. Run it against your own store.

Example numbers: a Shopify brand doing about $10M a year had 250,000 sessions over last year's BFCM window, Thanksgiving through Cyber Monday. It converted at 2.4% with an $85 average order.

  • BFCM revenue: 250,000 × 2.4% × $85 = $510,000.
  • AI-referred sessions last year: 0.6% of 250,000 = 1,500.
  • This year at +130%: 1,500 × 2.3 = 3,450 sessions.
  • At the site's 2.4% conversion: 3,450 × 2.4% × $85 = $7,038.
  • With a 40% conversion edge (3.36%): 3,450 × 3.36% × $85 = $9,853.
  • With a 48% edge (3.55%): 3,450 × 3.55% × $85 = about $10,400.

That is $7,000 to $10,400, or 1.4% to 2% of the example brand's BFCM revenue. It justifies an hour of setup before BFCM.

The count is also a floor: checkouts inside an agent, and shoppers who read an answer and come back later through direct or brand search, do not show up as AI referrals. And a wrong cutoff answer on December 18 turns into a late gift, a refund request and a support ticket.

Fill in the Agentic knowledge base before Thanksgiving

Write every entry as a full sentence with a date or a number in it, so an agent can quote it as written.

  • Holiday shipping cutoffs by method. Example: "Order by December 17 for standard delivery by December 24. Order by December 20 for express." Use your carrier and 3PL dates, not last year's.
  • Return window. How many days, whether the clock starts at order or delivery, who pays return shipping and how long refunds take.
  • Holiday returns. Example: "Orders placed November 1 through December 24 can be returned until January 31."
  • Gift policy. Gift receipts, gift messages and shipping to a different address.
  • Exclusions. Final sale items and BFCM bundles with different return rules.

Then make the knowledge base, your policy pages and your support team's saved replies match word for word. If an agent says 30 days and your refund page says 14, support ends up settling the difference.

Move your quiz answers into product data

Many DTC brands answer fit, shade and roast questions in a quiz or lower on the product page. An agent may never see either. It reads titles, attributes, variants and descriptions, so put the answers there:

  • Coffee: roast level, tasting notes, grind options, brew method and bag weight.
  • Cosmetics: shade range, undertone, skin type, finish and key ingredients.
  • Jewelry: metal, plated or solid, sizes, hypoallergenic or not, and resizing terms.
  • Apparel: fit notes such as runs small, fabric, inseam, and the size the model wears.

Fill in every variant. Start with last BFCM's top 20 products by revenue.

Set up an AI referral channel in GA4

Some guides report that Google added a default AI channel to GA4 in 2026. I could not confirm that in Google's documentation, so check your property's default channel group first. If there is no AI channel, build one:

  1. In GA4, go to Admin > Data display > Channel groups.
  2. Click Create new channel group and name it "Default plus AI." It starts as a copy of the default channels.
  3. Add a channel named "AI Assistants." Set the condition to Session source matches regex: `chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai`
  4. Drag AI Assistants above Referral. GA4 applies the first rule that matches, so placed below Referral it misses those visits.
  5. Save. In Reports > Acquisition > Traffic acquisition, switch the channel dimension to your new group.
  6. Set the dates to last year's BFCM window for a baseline, and compare conversion rate and average order with Organic Search and Paid Social.

Expect gaps. Some AI apps strip the referrer, so those visits land in Direct, and Google AI Mode clicks may show up as Google organic.

In Shopify, check the referrer reports for the same domains, and orders by sales channel for checkouts inside an agent, since those may never create a session on your site.

Check what the assistants say about your top products

Ask ChatGPT, Gemini, Perplexity and Copilot the same four questions about each of your top 5 products. If you sell on Amazon, ask Alexa for Shopping too.

  1. "How much is [product] and is shipping free?"
  2. "If I order [product] on December 18, will it arrive by Christmas?"
  3. "What is [brand]'s return policy?"
  4. "What is a good alternative to [product]?"

Score each answer right, wrong or missing. A wrong answer points to a gap in your knowledge base, product data or policy page. The fourth shows which competitors sit next to you.

That is 100 answers across five products and five assistants. Do your best seller in ChatGPT and Gemini first, and the rest before Thanksgiving.

Decide on Amazon's agent on purpose

Amazon renamed Rufus to Alexa for Shopping in the US in May 2026. According to this Alexa for Shopping and Rufus guide from amalytix, Sponsored Products ads inside it switch on automatically for existing campaigns, and Buy for Me lets Amazon's agent complete purchases on a brand's own site. Decide whether you want that, and watch what it does to your customer data and returns.

Run the BFCM pre-flight checklist

ItemWhereDone when
Holiday shipping cutoffsAgentic knowledge base, shipping pageDates match your carrier and 3PL
Return window and holiday returnsKnowledge base, refund page, support repliesSame wording in all three
Gift policyKnowledge baseReceipts, messages and ship-to covered
Agent checkout settingsSales channels > AgenticYou know which agents check out in chat
Product data, top 20 SKUsProductsAttributes and every variant filled in
AI Assistants channelGA4 channel groupsAbove Referral, last year's baseline pulled
Assistant spot checkChatGPT, Gemini, Alexa for ShoppingBest seller answers are right
Amazon Buy for MeAmazon accountDecision written down

Black Friday is November 27 and Cyber Monday is November 30. Aim to finish by November 20 so nothing changes during peak week.

Read the numbers in January before you change the plan

After Cyber Monday, compare the AI Assistants channel with your site average. Write the rules down now:

  • Under 1% of BFCM revenue: keep it to a monthly check of the knowledge base and the four questions.
  • Over 3% and converting at or above site average: give product data and the knowledge base an owner and a monthly review.
  • In between: check again after your January sale.

Add the channel as one line in your weekly report on new-customer MER and contribution margin, next to the numbers that set the budget. For where AI search and the other channels fit in a full plan, see the 2027 Marketing Impact Report for DTC brands.

Get help putting this into your operating rhythm

This work touches strategy, data, and the friction between a question and a checkout, three of the seven dimensions in the Marketing Impact OS. We install that system so checks like these keep running after the holidays.

I have helped four DTC brands in coffee, cosmetics, jewelry and apparel grow from four to six figures in revenue to seven to nine figures, with lower COGS and higher AOV. More about me and the team.

The free Marketing Impact assessment scores your brand on all seven. The Marketing Impact Roadmap starts from $10,000 and takes about four weeks. See how we work as a fractional CMO for DTC brands, or compare the top fractional CMOs for DTC brands.

Questions owners ask

What is agentic commerce on Shopify?

Agentic commerce is shopping where an AI assistant finds, compares and sometimes buys products for the shopper. On Shopify, the Agentic sales channel connects your catalog to assistants such as ChatGPT, Microsoft Copilot, Google AI Mode and the Gemini app, and the knowledge base sets what those agents say about your shipping and returns.

Can customers buy my Shopify products inside ChatGPT?

It depends on the agent and your settings. Coverage of Shopify's Fall '26 Edition says Shop Pay checkout works inside third-party AI assistants, but Shopify's own help pages have described ChatGPT as a referral channel where the shopper clicks through to your store to buy. Check the Agentic sales channel in your admin for what is live.

How do I track AI referral traffic on Shopify?

In GA4, create a custom channel group with an AI Assistants channel that matches sources such as chatgpt.com, perplexity.ai and gemini.google.com, and place it above Referral. In Shopify, check referrer reports and orders by sales channel. Some AI visits arrive with no referrer and land in Direct, so treat the count as a floor.

Does AI-referred traffic convert better than other traffic?

Adobe's data says it does, but write-ups disagree on the size of the edge, 40% higher in some and 48% in others. It is still a small share of traffic. Measure your own conversion rate and average order for the channel over BFCM before you move budget.

How do I get ChatGPT to recommend my products?

No setting guarantees a recommendation. Complete product data, specific policy answers in the knowledge base, and matching policy pages make your products easier for agents to read and compare. Then ask the assistants about your top products and fix what they get wrong.