AI Search Holiday Shopping Preparation, Week by Week

Black Friday is 27 November 2026, and the same sale reaches Google, ChatGPT and Shopify's agent catalog through different fields that switch on and off at different times. Below is the plan: which field does what, one product worked through a four-day sale, and what to do each week from 19 October so the price an assistant quotes is the price you charge.

Why start six weeks out?

Because pages are slow and feeds are not. A feed price is re-read with every upload or sync, so a wrong price can still be fixed late. A wrong fact on your own pages has to be crawled again before answers change, and a wrong fact on other sites takes longer still. Those are the fixes that need October.

For the site itself, my published timeline is: "Access fixes land in days. Content starts getting retrieved in weeks." The layer after that, what other sites say about you, shows consistent movement at three to six months, so nothing there will move before 27 November.

Adobe's forecast, as reported by Digital Commerce 360, expects AI traffic to US retail sites to rise 130% year over year this holiday season. It is a forecast of growth for US retail sites, not a share of traffic. If it holds, more of your holiday visitors arrive with a price an assistant already quoted them.

A wrong quote costs more than a missing one. On LuckyPaws, one of our own brands, ChatGPT returned service prices that did not match the published rates, and our field note puts it plainly: "A missing price is a gap; a confident wrong price is a figure a customer arrives with." A sale that ended on Monday but is still in an AI answer on Wednesday is that figure.

Which fields carry a sale, and when do they take effect?

Each surface reads a sale from its own field with its own start rule. Google can schedule a sale from an effective date, unless your feed comes from Shopify's app. OpenAI's file feed reads sale_price as the price right now. Shopify Catalog passes compare-at prices to agents only under three conditions. Promotions are a separate system on Google and OpenAI.

Google's sale price page says to send the effective date to mark when a sale applies, "otherwise the sale price will be used for your product immediately." The same page asks you to show the regular and the sale price on the landing page, and only the sale price at checkout.

Most Shopify stores do not send Google a feed they edit by hand. With the Google & YouTube app, Google's own guide says you set a compare-at price and a lower price in Shopify, and "prices will update on Google within 24 hours." There is no effective date in that route. Your Shopify price change is the schedule, and the lag after it is up to a day.

OpenAI's products spec accepts a sale date range in its file format, then says what it does with it: "Accepted as date metadata; does not schedule sale activation." Elsewhere the spec asks you to update the feed when a sale starts or ends. The documentation gives no time for how fast a snapshot reaches ChatGPT, so treat that gap as unknown.

The API route works differently. In OpenAI's products API, price is the "Active sale price for this variant" and list_price is the reference price before any discount, in minor units. The promotions endpoint takes a start and end time and three kinds of benefit: an amount off, a percent off and free shipping. It describes the data, not where ChatGPT shows a promotion. Our ChatGPT product feed guide covers who can use which route.

On Shopify, the agent side changed on 9 October 2026. Shopify's changelog says Catalog now gives agents your compare-at price as list_price, so they can tell a product is marked down. A variant carries it only when compare-at sharing is on in Catalog Mapping, the compare-at price is higher than the price, and you show compare-at prices in the buyer's region. As the changelog puts it, "Shopify turns this on by default for most stores and turns it off for stores that use compare-at for MSRP or another reference price." Our Shopify Agentic Storefronts guide explains where Catalog sends your products.

One kettle through a four-day sale

Take a pour-over coffee kettle sold at €80 in the EU and $96 in the US, discounted 25% from Friday 27 November to Monday 30 November. That is €60 and $72 for a four-day sale. Each surface needs the markdown written its own way, and in the EU the reference price depends on the previous 30 days.

Shopify. Set the price to €60 and the compare-at price to €80 at the start, and restore them at the end. This one change feeds three surfaces: your store, Google through the app, and agents through Catalog. For EEA buyers, Catalog passes the €80 to agents only if you show compare-at prices to them.

Google Merchant Center. Through the Shopify app, Google can show the old price for up to a day after the start and the sale price for up to a day after the end. The first lag is harmless, because a shopper pays less than Google showed. The second is the risky one, so check a few products in Merchant Center on 1 December and fix any that still show €60. If you control your own Google feed, keep price at 80.00 EUR, add sale_price 60.00 EUR and sale_price_effective_date 2026-11-27T00:00+0100/2026-11-30T23:59+0100, upload it in the week of 16 November, and change the store price at the same minutes so feed and page agree.

OpenAI file feed. Write price 96.00 USD and sale_price 72.00 USD in a snapshot you upload early on 27 November. If ChatGPT still shows $96 for the first hours, a shopper pays less than they were told, which does no harm. The end is where it goes wrong: upload a sale-off snapshot without sale_price in the last hours of 30 November, US time. If nobody can do that, honor $72 for orders that arrive quoting it until ChatGPT shows $96 again.

OpenAI products API. If you push the sale through the API instead, send price 7200 and list_price 9600, currency USD, at the start, and price 9600 without the sale at the end.

Google promotions, if you add a code. Say you also offer free shipping on orders over €100 with a code. That belongs in a Merchant promotion, and Google's FAQ says review "is typically completed within 24 hours of the promotion effective start time, but can take up to 72 hours." It also recommends submitting at least 24 hours before the start. Even on time, approval can land up to three days into a four-day sale. Keep the discount itself in the price fields, which do not wait for promotion review, and let the promotion carry only the extra.

The plan, week by week

Six weeks is enough if fixes come in the right order: find wrong facts first, fix the pages they came from while there is time for a new crawl, then set prices once and freeze them. The dates assume a sale from 27 to 30 November 2026.

1. Week of 19 October: record what AI says today. Write ten questions a holiday buyer would ask, such as "best pour-over kettle under $100 as a gift", run them in ChatGPT, Google AI Mode, Perplexity and Gemini, and log every wrong price, product, policy or shipping claim. Our measurement protocol keeps an error register for exactly this, and that register, "not the score, is usually what a client acts on first."
2. Week of 26 October: fix the pages the errors came from. Prices and specs in the server-rendered HTML, a holiday returns page and shipping cutoffs on a page crawlers can reach, and product copy that matches your feed. In the EU, keep the regular price of anything you will discount steady from 28 October, or plan for the lower reference.
3. Week of 2 November: clean the data you control. In Shopify, open Catalog Mapping and decide on compare-at sharing; remove any compare-at value that was never a real price. In Merchant Center, fix disapproved products now.
4. Week of 9 November: build the sale, do not switch it on. Prepare the Shopify price and compare-at changes as one bulk edit, or in the scheduling app you already use, or prepare your own Google rows with effective dates. If you run a direct OpenAI feed, prepare two snapshots, sale on and sale off, and name who uploads each and when.
5. Week of 16 November: submit and upload. Send any Merchant promotion by Friday 20 November so the review has working days to run. Upload your own scheduled Google sale prices if you use them.
6. Week of 23 November: freeze and check. No price or copy changes after Tuesday 24 November. Re-run your ten questions and compare them with week one. Check that stock changes reach Google and Catalog within the day, and for a direct OpenAI feed, plan an extra snapshot during the sale for items likely to sell out, with availability updated.
7. 27 to 30 November: run it. Upload the sale-on OpenAI snapshot early on the 27th and check one product on every surface. On the evening of the 30th, upload the sale-off snapshot, and restore the Shopify prices at 23:59.
8. 1 December: check everything. Look for sale prices that stuck in Merchant Center and ChatGPT, and run your questions again within the week.

What the ranking guides leave out

The pages ranking for this query agree on the basics: be crawlable, collect reviews, add schema, start early. Of the eight we read on 17 October 2026, none gave a dated plan for getting a sale price right in Google, ChatGPT and Shopify Catalog. The one dated playbook, Alhena's, is for an on-site chatbot.

Six weeks will not build AI visibility. It can correct what engines get wrong about you, which is the lesson our own field notes reached: "Presence was never the issue. Accuracy was."

What to do on Monday

Run the week-one step first: ten buyer questions, four engines, every error written down with the engine and the date. That list decides the next five weeks: which pages to fix in late October, which prices and policies to check in November, and what to look for again on 1 December.

One run per engine is a snapshot, and answers vary from run to run. If you want repeated runs and a logged error register, that is where our AI visibility audit starts.