As of November 12, 2026, the sharpest number in Adobe's holiday forecast is not the $275.1 billion of projected US online spend but the 130 percent growth it expects for AI-referred traffic to retail sites. That figure sits on top of a 693.4 percent jump measured in the 2025 season, which means the channel is doubling again from a base that has already quadrupled. For e-commerce teams, the practical question is narrower than the headline: what has to be true of a product listing before an assistant will name it at all.
What Adobe projects, and what the projection leaves out
The forecast covers November 1 to December 31 and draws on Adobe Analytics data from more than a trillion visits to US retail sites, 100 million SKUs and 18 product categories. Online spend is projected at $275.1 billion, up 6.7 percent, against $257.8 billion and 6.8 percent growth in 2025. Cyber Week accounts for $47.5 billion, or 17.3 percent of the season, with Black Friday at $12.9 billion and Cyber Monday at $15.1 billion.
Adobe counts AI traffic as shoppers clicking a link from AI-powered chat and browser services. That definition matters: it excludes answers where a brand is named but no link is clicked, which is the outcome most assistant interactions now produce. A 130 percent rise therefore describes referral clicks, not total exposure, and it says nothing about the share of retail visits AI already controls.
The referral is not the click, and the click is not the sale
The 2025 comparison is the useful part of the release. AI traffic rose 693.4 percent across the season and 670 percent on Cyber Monday, and Adobe still described the user base as modest. A fall to 130 percent growth is what a channel looks like after it has multiplied several times over an existing footprint, not evidence of a plateau.
Why the base matters more than the rate
Growth rates on small bases mislead budgets. If AI referrals were a rounding error in 2024, a 693 percent rise still leaves a channel that most retailers cannot size from their own analytics. The 2026 forecast is more actionable because it arrives alongside an absolute spend pool, which lets a merchant test whether referral sessions convert at, above or below the site average instead of arguing about market share.
| Metric (US online) | 2026 forecast | 2025 actual |
|---|---|---|
| Holiday season spend | $275.1 billion, up 6.7 percent | $257.8 billion, up 6.8 percent |
| Cyber Week spend | $47.5 billion, up 7.4 percent | $44.2 billion, up 7.7 percent |
| Black Friday | $12.9 billion, up 9.2 percent | $11.8 billion, up 9.1 percent |
| Cyber Monday | $15.1 billion, up 6.2 percent | $14.25 billion, up 7.1 percent |
| AI traffic to retail sites | Up 130 percent | Up 693.4 percent |
Read together, the two columns separate demand from distribution. Online spend grows at a stable rate, while the route shoppers take to a listing is being rebuilt. A retailer that plans inventory against the first column and ignores the second will hold the right stock and still lose the query.
What a product listing has to carry now
Attributes that survive extraction
Assistants do not browse a storefront the way a shopper does; they pull attributes and assemble an answer. A listing whose title carries the selling point but whose attributes are empty gives the model nothing to compare. Category growth in this environment depends on how completely a listing describes size, material, compatibility and price conditions, because those are the fields a comparison answer is built from.
Access, feeds and measurement
Three technical items sit upstream of any content work. Crawler access decides whether an assistant can read the page at all, and blocked crawlers produce no error message. Product feeds decide whether the listing enters a shopping surface with the right price and availability. Measurement decides whether the referral is visible in reporting: many AI clicks arrive without a referrer tag, so server-side tagging and a separate referral segment are the only way to see the channel.
Consumer demand is shifting along the same line. Adobe's July 2026 survey of 5,000 consumers found that 77 percent of shoppers who used AI for online shopping felt more confident in the purchase, and 69 percent said they were less likely to return an item. Confidence changes the return rate, and the return rate changes the unit economics of a listing that was already thin on margin.
Discount depth meets a credibility problem
Adobe's forecast puts peak discounts at up to 14 percent in early November, up to 21 percent just before Thanksgiving and up to 30 percent during Cyber Week, measured against listed prices. Those are seasonal upper bounds, not averages, and they apply unevenly across categories.
Consumer research points the other way. In a Drive Research survey of 1,025 Black Friday shoppers, 78 percent said they take part for the discounts, while 64 percent said they believe retailers raise prices beforehand to exaggerate savings. The second number records a belief rather than a finding of manipulation, but it changes what a listing has to prove. A discount badge without a reference price, a threshold and an end date answers none of the questions a sceptical shopper is actually asking.
Delivery terms carry similar weight. In the same survey, 74 percent said free shipping influences where they shop, 47 percent valued guaranteed arrival before the holidays and 43 percent valued fast shipping. These are stated preferences, not purchase records, but they set the order in which a listing is read: price, then delivery cost, then arrival date.
Four checks before Thanksgiving
Most of the work that decides whether an assistant can use a listing is unglamorous and finishes before the traffic arrives. The sequence below is ordered by how quietly each item fails when it is wrong. None of the four costs much on its own; what they cost is the season, if nobody checks them before the traffic lands.
- Confirm crawler access in robots.txt and at the firewall, and re-test after any change, since propagation can take up to a day.
- Sync the product feed so price, availability and variant attributes match the page a shopper lands on.
- Complete the attribute set for the SKUs that carry the category, rather than for the whole catalogue at once.
- Tag AI referrals server-side and report them as a separate segment, so the channel is measured instead of inferred.
Run the four against a sample of twenty listings and the failure pattern is usually obvious within an hour. The table below is the minimum record worth keeping for each SKU in that sample. Keep the sample small enough to re-run in November and again in December.
Keep the record for the season and the comparison across years becomes possible, which is the only way to tell whether a listing improved or the market simply grew underneath it. Without that baseline every December looks like progress, and the same four checks get repeated next year with no evidence that any of them mattered.
FAQ
Does a 130 percent rise in AI traffic mean AI drives most retail visits?
No. It is a growth rate on a small base, and Adobe does not publish AI's share of visits or sales.
Why count AI referrals by clicks rather than by mentions?
Because clicks are the only part a retailer can observe directly. Mentions inside an answer leave no trace in analytics.
Which listing change moves AI visibility fastest?
Filling in the attributes a comparison answer needs, since empty fields give a model nothing to quote or rank.
Are the forecast discounts averages?
No. They are seasonal upper bounds by period, applied unevenly across categories and merchants.
