Holiday 2026: preparing for AI shopping questions before Black Friday

Holiday shopping answers depend on boring facts: a price that matches on every page, a stock status that is true today, a shipping date that still holds. You cannot decide which products an assistant picks. You can decide whether the facts it reads about yours are current. This checklist runs backward from Black Friday (Nov 27) and Cyber Monday (Nov 30), then forward to the December shipping cut-offs.

What do the holiday forecasts say, and what do they leave out?

Thanksgiving is the fourth Thursday of November in federal law, which makes it Nov 26 this year. Black Friday follows on Nov 27, and Adobe places Cyber Monday on Nov 30.

Two dated forecasts are worth knowing:

• Adobe (Sept 28, 2026): $275.1 billion of U.S. online holiday sales from Nov 1 to Dec 31, up 6.7% on last year, with $15.1 billion expected on Cyber Monday. Adobe also forecasts that AI traffic to U.S. retail sites will rise 130% on last year, and says it measures that traffic by shoppers clicking on a link.
• NRF (Oct 15, 2026): retail sales in November and December up 4.2% on 2025, to $1.013 trillion. NRF's figure covers core retail and leaves out auto dealers, gas stations and restaurants.

Neither forecast tells you how often an assistant will recommend your brand. Adobe's AI figure counts clicks, and Dopbot's index counts answers. The October 2026 edition was collected starting Oct 23, before the season, with five questions per category (the fifth asked for an alternative to a named brand). Read it as a pre-season picture of how assistants answer, not a holiday forecast.

Where might the facts in an answer come from?

As of October 2026, OpenAI's help center says ChatGPT's shopping listings carry prices it receives from third parties, and warns that a merchant's price or shipping change can take a while to appear. It adds that Shopify merchants' product data already reaches ChatGPT through Shopify Catalog. As of October 2026, Perplexity's help center says merchants that supply fuller product details, such as stock, reviews and pricing, are more likely to be recommended.

Our October sample shows which pages appear in each engine's source lists:

Both rows count all of each engine's answers in the October 2026 edition (United States, Mexico and Canada, all five questions, including the alternative-to-a-brand question), not only U.S. answers. An own-site match is a domain whose first label equals a recommended brand's name (samsung.com for Samsung); it misses country domains, parent companies and other store names. The retailer list is hand-made. For Perplexity the lists are pages its search retrieved, not pages the answer necessarily uses.

The practical reading: your own product pages and the retailer pages that sell your products are both within reach of these answers, so both need the right price and stock. A domain in a source list does not show that it shaped a recommendation, and we did not measure whether any answer's price was correct.

What should you do, and when?

Oct 30 to Nov 5: baseline and audit

• List the products you most want recommended in holiday and gift answers, and five question types for each: best overall, budget, premium, a specific use, and an open buying-advice question (the index's November design uses a buying-advice question as its fifth).
• Record a baseline. In fresh sessions, note whether you are recommended and what price and stock the answer states. Ask each question 20 times per assistant if you want a reading per question (100 answers per assistant for one product); the free manual method asks each of its five questions 4 times, which gives 20 answers per assistant read together, and gives a sheet layout.
• Write one fact sheet per product: current price, sale price and its dates, availability, ship-by dates, return window, GTIN or MPN, and the exact brand name.
• Check crawler access. OpenAI says a site is eligible for ChatGPT search results when OAI-SearchBot can crawl it and the host or content delivery network allows OpenAI's published searchbot IP addresses. It adds that placement is not guaranteed.

Nov 6 to Nov 12: make every source agree

• Page, markup, feed and retailer listings should say the same thing. Google's merchant listing guide requires the active price in offer markup, with a priceCurrency, and recommends putting Product markup in the initial HTML. Dynamically generated markup, Google says, can make Shopping crawls less frequent and less reliable.
• Schedule feed updates to match store changes. Google names the time difference between a website update and a Merchant Center update as a common cause of wrong availability. Its automatic item updates are meant for sporadic problems and are not a replacement for regular product data updates.
• Set the holiday return window. As of October 2026, Google lets you extend a return window for a date range, with a number of days and a no-later-than date. Put the same terms on your returns page.
• If you sell on Shopify, open the Agentic section of the admin. As of October 2026, it has listing quality indicators that check description length, images, reviews, variants with in-stock availability, and shop policies, and a search preview that shows how a product might rank, which Shopify calls a directional signal.
• If you do not sell on Shopify or Etsy, OpenAI's merchant page says merchants can apply to share a product feed and that a self-serve merchant platform is still being built; the page is undated and gives no firm date. As of October 2026, its developer documentation says feed onboarding is open to approved partners. The feed specification asks you to update availability when an item sells out and to update the feed when a sale starts or ends.

Nov 13 to Nov 19: retailers and editorial pages

• Check every retailer and marketplace that sells you, for price, stock and the brand name on the listing, and send corrections to the retailer.
• Look at the review pages in your category's source lists on the index category pages (candles is one of the 40 updated US categories, so its page shows the latest update). In the October 2026 edition, a Wirecutter page was among the retrieved sources of 333 of 787 Perplexity answers, and none of the 800 OpenAI answers linked nytimes.com.
• Ask editorial sites for their holiday deadlines instead of assuming them. We could not verify any publisher's lead times.

Nov 20 to Nov 26: freeze and re-measure

• Freeze the product page templates, feed mappings and sale schedules.
• Schedule the feed update for each sale's start and end, since OpenAI's feed specification asks for the current price and an update when a sale starts or ends, and confirm stock buffers for the products you expect to sell fastest.
• Repeat the baseline questions so you have a pre-sale reading to compare with.

Nov 27 to Dec 3: Black Friday to Cyber Monday

• Ask your top questions each day, a handful of times per assistant. This catches wrong prices and out-of-stock claims; it is too few answers to estimate share.
• When an answer is wrong, record the exact prompt, assistant and time, open the sources it shows, and fix the one that is stale.
• OpenAI's help center invites a thumbs-down or a report from the menu on a product with incorrect information, and tells shoppers to confirm final price, availability and returns with the merchant.

Dec 4 to Dec 17: shipping cut-offs

• USPS recommends sending by Dec 17 for Ground Advantage and First-Class Mail, Dec 18 for Priority Mail and Dec 19 for Priority Mail Express, for delivery before Dec 25 in the lower 48 states. USPS calls these recommendations and says actual delivery dates can vary.
• Put your own last-order dates on product pages and in the feed. Shipping terms are among the facts OpenAI says can lag behind a change.

After the season

Ask the baseline questions again and compare, using the intervals below.

How should you read the before and after numbers?

Each answer is one draw, so small counts move a lot. The 95% Wilson intervals for a brand recommended in 25% of answers:

At 20 answers, 5 of 20 and 10 of 20 have overlapping intervals, and an exact test between them gives p of about 0.19, so a move that size is not proof of change. The post on how many answers to trust explains the test. From the Nov 2 run, the index's 40 updated US categories use 40 answers per category for the same reason; the October 2026 edition has about 10. Keep the same questions, assistants and settings for both readings, and remember that API answers differ from app answers, as the API vs app post explains.

Frequently asked questions

Can I put my products into ChatGPT's shopping results before the holidays?

If you sell through Shopify or Etsy, OpenAI says your catalog is already integrated and no application is needed. Other merchants can apply to share a feed, but as of October 2026 OpenAI's developer documentation says feed onboarding is open to approved partners, and its merchant page says a self-serve platform is still being built, so do not plan around being live before Black Friday. OpenAI says shopping is live for ChatGPT users in the U.S. as of October 2026.

Does the AI traffic forecast mean assistants will send me shoppers?

Not necessarily to you. Adobe forecasts that AI traffic, measured by clicks, to U.S. retail sites overall will rise 130% on last year. It does not say which brands those clicks will reach. Measure your own visibility and traffic, and see the guide to AI visibility basics.