If I ran a small shop, I'd spend this week making my product pages easy for an AI assistant to read before I spend another cent on ads. That sounds like a strange priority with Cyber Week starting tomorrow, so let me show you the numbers that pushed me there.
On 19 November Salesforce published its Cyber Week forecast, and the headline figure was not the total. It expects AI and agents to influence $75.8 billion of global online sales during Cyber Week, which it puts at 21% of a record $357 billion and up 13% from $67 billion in 2025. In the same release it says AI-driven traffic to retailer sites is up 104% year over year, and that shoppers who arrive from an AI referral are two times as likely to convert as typical ecommerce shoppers. Cyber Week, in Salesforce's definition, runs from Tuesday 24 November to Monday 30 November.
Put plainly, a growing share of your customers will ask ChatGPT, Gemini or Google's AI Mode what to buy, read a short answer, and click through to whichever shop the assistant named. The ones who do click are close to buying. If the assistant can't work out your price, your sizes or your returns policy, it will name somebody else.
What the data says, and what it doesn't
These are big, round, vendor-published numbers, so they deserve a careful read before you change your week around them.
First, the $75.8 billion is a forecast, not a measurement. Salesforce says its predictions come from its Shopping Index, which it describes as the activity of more than 1.5 billion shoppers across 1.5 trillion page views, built from aggregated data across its own Agentforce, Commerce Cloud, Service Cloud and Marketing Cloud products in more than 89 countries. It also says the calculations "blend together first-party and third-party data, as well as several market assumptions" and are extrapolated to the wider retail industry. In other words, the base is retailers that run on Salesforce's software, scaled up with modelling, and that group looks nothing like a two-person shop.
Second, Salesforce does not define "influence" on that page. The closest it has come is in an earlier Cyber Week release, where it credited AI and agents with "personalized recommendations, targeted promotions, and smarter customer service". So the $75.8 billion almost certainly includes the "you might also like" carousel on a retailer's own site, not only shoppers who came from ChatGPT. I'd treat it as a measure of how much selling now passes through some kind of AI, rather than a measure of chatbot shopping.
Third, the conversion figure has an obvious selection effect. Someone who has already asked an assistant "best waterproof hiking boots under $150" and clicked one of three answers is further along than someone browsing from an Instagram ad. Twice the conversion rate is partly the assistant's doing and partly the shopper's intent. That doesn't make it less useful to you, because those are exactly the shoppers you want, but it means you shouldn't expect any visitor to convert twice as well just because a chatbot sent them.
The direction is backed up independently. Adobe, using its own data on more than 1 trillion visits to US retail sites, forecast in September that AI traffic to US retail sites would rise 130% year over year this holiday season, with the largest jump, 159%, on Thanksgiving. Digital Commerce 360 reported that Adobe saw AI referrals bring in 83% more revenue per visit than non-AI traffic in September. Two firms with different data sets point the same way.
Why it is happening now
Shoppers have changed where they start. Salesforce says one in five shoppers will start their holiday shopping with AI, and that the top use for personal agents is price comparison. Comparison is tedious work, and assistants do it in seconds, so people hand it over.
The platforms are making it easier. OpenAI's merchant page says shopping is live for ChatGPT users in the US, that merchants can send product feeds through the systems they already use, and that Shopify and Etsy catalogues are already integrated with no application needed. Google already reads product structured data and Merchant Center feeds for its shopping results. Retailers are adding their own assistants too: Salesforce expects one in three brands to have shopper agents on their websites by Cyber Week, and says shoppers who use one spend 8% more per order and buy 23% more items (again, people who chat with a store's assistant are probably keener to begin with).
Trust still stops at the till. Salesforce found 73% of shoppers aren't ready to give a personal agent their card details, and only 8% have let one make a purchase. For now, the assistant recommends and the human buys on your site. That is good news for a small shop, because the sale still happens on a page you control.
What it means for you
For a small business, the risk is quiet. You won't see a drop in a dashboard when an assistant skips you; you simply won't be in the answer. Assistants work from what they can read, so a product page whose price sits inside an image, whose sizes live in a pop-up, and whose returns policy is a PDF gives them very little to go on. A competitor with plainer pages gets named instead.
The cost of fixing it is mostly time. Writing clear specs, adding structured data and connecting a free product feed costs nothing but hours, which compares well with paid ads at Cyber Week, when everyone is bidding on the same clicks. For creators who sell merch or digital products, the same applies to a Shopify, Etsy or Gumroad listing. For indie builders, there is a services market here, which I'll come to.
I wrote earlier this year that AI won't overtake Google search in 2026, and I still think that's right. Most of your traffic will keep coming from search, social and email. But the AI share is small, growing fast and unusually ready to buy, and the work that makes you readable to assistants also helps Google's ordinary shopping results.
A worked example (illustrative maths)
The figures below are invented to show the shape of the decision. They are not benchmarks, and your numbers will differ.
Take a small online shop selling leather goods: 6,000 visits in Cyber Week, a 2% conversion rate and an average order of $60. That's 120 orders and $7,200. Say 3% of visits, or 180, come from AI assistants. If those visitors convert at twice the shop's average, as Salesforce's data suggests is possible, they produce about 7 orders, or roughly $430.
Now suppose the shop's pages are hard to read, and the assistants name it half as often. It loses about 90 of those visits and around 3 or 4 orders, roughly $215. That's small this week. But if AI referral traffic roughly doubles again over the next year, as the 104% figure suggests it can, the same gap costs twice as much next Cyber Week, and every other week in between.
Compare that with ads. At an assumed $1.20 per click, $215 buys about 180 paid visits, which at 2% conversion gives around 3.6 orders, and the spend is gone when the campaign stops. The page work costs a day or two and keeps paying while the pages stay accurate. That trade-off is why I'd do the pages first.
What to do this week
Ask the assistants about yourself. Type the questions your customers would ask ("best leather card holder under $80 in Australia", "is [your shop] legit", "[your product] vs [competitor]") into ChatGPT, Gemini and Perplexity. Write down whether you appear, what they say your price is, and anything they get wrong. Repeat monthly.
Put the facts in plain text. Price, currency, sizes, dimensions, materials, compatibility, what's in the box, and who the product is for should all be readable words on the page, not text baked into images or hidden in tabs that load later.
Answer the comparison question. Add two or three sentences on how the product differs from the obvious alternative, and a short FAQ with real answers. Price comparison is the top job people give personal agents, so give them something to compare.
State shipping and returns on every product page. Delivery times, costs, free-shipping thresholds, return window and who pays return postage, in one or two sentences each.
Add product structured data. Google's merchant listing structured data requires a name, image and offer, and recommends brand, GTIN, ratings and more. Its return policy markup can sit once at organisation level. Most Shopify themes and WooCommerce plugins output some of this already; check it with the Rich Results Test and fill the gaps.
Connect your product feed. Set up Google Merchant Center if you haven't, and enter shipping and returns there or in Search Console. If you sell in the US on Shopify or Etsy, your catalogue is already connected to ChatGPT shopping; other US merchants can apply on OpenAI's merchant page, which currently runs a waitlist.
Label AI traffic in your analytics. Create a channel group for referrers such as chatgpt.com, perplexity.ai and gemini.google.com so you can see whether the work moves anything.
If you want a quicker read on where you stand, Dedica checks whether AI agents can find, read and buy from your site. And if your holiday plan leans on your own posts, as it does for 45% of small businesses, link those posts to the product pages you've just cleaned up.
The money angle for builders and freelancers
Plenty of small retailers know AI is changing shopping without knowing what to change on their own site, and that gap is a service you can sell. An "AI shopping readiness" audit covers the steps above for a store's top 20 products: what three assistants say about the shop, missing or broken structured data, facts locked inside images, gaps in shipping and returns information, and whether a Merchant Center feed exists. Deliver it as a short report with fixes ranked by effort.
My estimate is that a fixed fee somewhere between $300 and $800 for the audit is reasonable for a small store, with the fixes billed separately. That's a guess, not market data, so test your price on the first few clients. The checks take an hour or two per store, the report is mostly repeatable, and the follow-on work (rewriting product copy, fixing theme markup, setting up feeds, building a simple on-site assistant) is where the larger invoices sit. I'd open each pitch with what the assistants currently say about the shop, because showing a shop owner that ChatGPT quotes their old price is more persuasive than any pitch deck.
A note of caution before you sell it: nobody can promise that an assistant will recommend a given shop, and the platforms change how they choose products without notice. Sell clarity and accuracy, report what the assistants said before and after, and leave rankings out of the promise.
Cyber Week starts on Tuesday, and a few hours today on your ten best-selling product pages is probably the most useful thing you can do before then, with the rest of the list to follow over the coming weeks.
Sources
Salesforce Data: Record Cyber Week Expected to Deliver $75.8B in AI-Driven Sales, Salesforce, 19 November 2026
Salesforce Data: AI Steers the Cyber Week Shopping Cart to $61 Billion in Sales, Salesforce, 20 November 2024
Adobe: U.S. Holiday Shopping Season to Hit Record $275.1 Billion Online, Rising 6.7% YoY, Adobe, 28 September 2026
3 online sales factors to watch in 2026 holiday shopping forecasts, Digital Commerce 360, 19 November 2026
Power product discovery in ChatGPT, OpenAI, accessed 22 November 2026
Merchant listing structured data, Google Search Central, accessed 22 November 2026
Merchant return policy (MerchantReturnPolicy) structured data, Google Search Central, updated 8 September 2026

