Adobe forecasts that traffic from AI chat and browser services to US retail sites will grow 130 percent year on year over the November and December holiday season, rising to 141 percent on Thanksgiving Day itself. Topline online spending is forecast to grow 6.7 percent, which Adobe puts at 275.1 billion dollars across the two months.
Every figure in the report measures US retail sites. Nothing in it covers Europe, which makes the 130 percent a direction worth understanding and not a number a European operator can plan against.
A growth rate without a base
The 130 percent is the most quoted figure of the forecast and the least examined. It is a growth rate, and Adobe does not publish the absolute volume of AI-referred traffic underneath it.
That matters more than it sounds. A channel can more than double while remaining a rounding error, and it can more than double because the comparison period barely existed. Assistant-referred shopping traffic was close to new a year ago, so a 130 percent rise is entirely consistent with a channel that is still small enough to ignore and entirely consistent with one that is not. The report does not let you tell which.
The honest reading is that the direction is established and the magnitude is not. Anybody converting 130 percent into a budget line is supplying the missing denominator themselves.
From the publisher
Everything we know about running Amazon, in one course.
The commercial strategy, taught by operators who run our own brands.
The discount curve is the usable part
The forecast’s most practical content is the shape of discounting rather than its depth. Adobe expects 14 percent off listed price in early November, 21 percent just before Thanksgiving and 30 percent during Cyber Week.
That is a schedule, and it is the sort of thing an operator can actually price against: a defined ramp with a known peak, which tells you what a shopper will be comparing your price to in each week of the season. Adobe also expects a Black Friday’s worth of spending to land in October, which pulls the start of the season earlier than the promotional calendar usually assumes.
Buy now pay later is forecast at 21.3 billion dollars across November and December. For a seller that is a working-capital and returns observation rather than a demand one, because deferred payment shifts both the basket size and the return profile of the orders it touches.
Volume up a little, discounts up a lot
Put the two headline numbers next to each other. Spending grows 6.7 percent and peak discounting reaches 30 percent off listed price. A season where volume rises modestly while promotional depth sits at nearly a third is a season decided on margin rather than on demand.
That is the squeeze, and it is the part of the forecast that reads the same on either side of the Atlantic even though the figures do not. A seller who plans to meet a 30 percent benchmark across a full catalogue is planning to buy revenue. A seller who picks the products where depth converts and holds price everywhere else is planning a season.
The mobile reading supports the same conclusion. Adobe expects 4.4 billion dollars spent on mobile devices on Thanksgiving Day, up 11 percent on 2025, with mobile taking 63 percent of that day’s spending. Peak days are phone days, and a phone shopper comparing two prices is doing it faster and with less patience for a slow listing.
What the method does and does not capture
Adobe Analytics builds the forecast from direct online transactions rather than from consumer surveys, across more than a trillion visits to US retail sites, 100 million SKUs and 18 product categories. That is a strong base for a demand forecast and it carries a structural limit worth naming.
It measures what happens on retail websites instrumented with Adobe’s software. Spending that never touches such a site, inside a marketplace app, inside a social commerce checkout or through an assistant that completes a purchase without a site visit, is not what this instrument is pointed at. As more of a season moves into those surfaces, a site-based measurement describes a shrinking share of it.
That is also why the AI traffic figure is measured as referrals to retail sites. It counts assistants that send a shopper somewhere. It does not count an assistant that answers and transacts without the shopper arriving anywhere, which is the version of this that platforms have been building towards all year.
For the business operator
If you sell into the United States, take the discount curve as a planning input and nothing else from this as a forecast. Fourteen percent in early November, 21 percent before Thanksgiving, 30 percent in Cyber Week is the benchmark your price will be read against, and knowing the ramp is worth more than knowing the season total.
If you sell in Europe, do not put the 6.7 percent in a European plan. It is a US figure from a US instrument, and the European season has a cost change landing in the middle of it that has no American equivalent. Use the AI referral trend as a reason to check one thing: whether your own analytics can currently identify assistant-referred traffic at all. Most cannot, and that is a configuration job rather than a budget one.
Separate the products where depth converts from the ones where it only costs margin, before the first week of November rather than during it. A 30 percent benchmark applied across a catalogue is the single most expensive default decision available this quarter.
From the consumer’s side
A shopper who has been trained by three years of deepening discounts does not treat a November price as a price. They treat it as an opening position, and they wait, because waiting has paid every year. That is what a published 30 percent Cyber Week expectation does: it tells the shopper what to hold out for.
The assistant changes the waiting. Asking a chat service to find the best price collapses a comparison that used to take several tabs and a quarter of an hour into one question, which removes the friction that previously made shoppers settle for the first acceptable offer. A seller whose strategy relies on being found first rather than on being the better offer is relying on friction that is being removed.
The three-line version
If you run peak trading, use the discount ramp and ignore the season total unless you sell into the United States. The line to take upward is that volume is forecast up 6.7 percent while peak discounting reaches 30 percent, so the season is a margin problem rather than a demand one.
If you have just taken on digital or demand planning, the gap this exposes is measurement. Assistant-referred traffic is forecast to more than double and most analytics configurations cannot separate it from direct or referral traffic, so you will not be able to tell afterwards whether it mattered.
If you sign the budget, nothing here justifies European spend. These are US figures from a US instrument, the 130 percent is a growth rate with no published base, and the season total is one firm’s forecast. Refuse any European plan that cites the 6.7 percent as a market growth assumption.
Sources
Adobe, Holiday shopping report: https://business.adobe.com/resources/holiday-shopping-report.html
PPC Land, Adobe puts US online holiday spend at $275.1 billion, up 6.7%: https://ppc.land/adobe-puts-us-online-holiday-spend-at-275-1-billion-up-6-7/