There are, roughly, two kinds of people reading Adobe's Prime Day Forecast this week.
The first kind will have it turned into a tidy carousel by Thursday: five slides, a "what this means for brands" header, a confident closing line about the future of retail. Genuine LinkedIn gyaan, dispensed to anyone scrolling past. The second kind actually believes AI is rewiring how people shop, reads a forecast like this, and immediately starts poking at the numbers that do not quite sit right.
I am, fairly unapologetically, in the second camp. So is everyone at ReFiBuy. I will let you guess which group I find more useful to spend time with.
So instead of the recap you have probably seen five times already, here are three things in the report I cannot stop turning over. The top line, for the record, is real: a record $26.3 billion in U.S. online spend across the four-day event, up 9% year over year, more than Cyber Monday and Black Friday 2025 combined. Good number. Not the interesting part.
Adobe expects traffic to retail sites from generative AI sources to grow about 103% during Prime Day versus June last year. A clean doubling. Lovely on its own, until you put it next to the recent trend: that same traffic was up 693.4% during the 2025 holidays, then 138% year over year in May 2026, and now a forecast 103% for this event.
Read literally, the growth rate is slowing, and I would bet money someone publishes the "is agentic commerce finally cooling off?" take within the week. I think that read is wrong, and I will happily be told so in the comments.
Here is my problem with it. Adobe is measuring one very specific thing: a shopper clicking a link from an AI tool through to a retail site. That is the visible, countable sliver of AI-assisted shopping. But the better these agents get, the more of the research and comparison happens inside the assistant, and the shopper lands on the purchase through a direct visit or a branded search with no AI referral attached. The influence absolutely happened. The attribution just evaporated.
So a slowing click-through number is completely consistent with rising agentic influence. Maybe I am too close to this to be objective, but when I see that headline AI figure flatten, I do not read it as agents mattering less. I read it as agents mattering in all the places our analytics cannot see yet.
I promised you lunch boxes, so here we are, and I mean this fairly seriously.
Most write-ups will quote Adobe's top-line discount depths and move on. The number I keep staring at is buried further down, in the category-level demand spikes. Adobe projects kids' apparel up 230% against the June daily average, lunch boxes up 215%, backpacks up 185%, car seats up 155%, portable chargers up 105%. These are not categories. They are oddly specific products, bought for oddly specific reasons: parents getting a jump on back-to-school, families packing for summer travel.
That specificity is the whole point, and I might be reading more into a lunch box than any one lunch box can hold. But this is exactly how AI shopping agents query a catalog. An agent does not go looking for "electronics." It works a precise intent: a portable charger that fits in a carry-on and can actually charge a laptop, a car seat that clears a specific safety standard and fits a specific back seat. The match happens attribute by attribute, SKU by SKU.
And here is the gap that keeps me up at night, professionally speaking. Most brands still publish at the brand and category level, gorgeous storytelling about the line, thin and inconsistent data about the individual product. Agents need the reverse: complete, structured, attribute-level data they can interpret with confidence. When that data is not there, the agent does not politely rank you a little lower for "best insulated lunch box for a long commute." It leaves you out of the answer entirely, and nothing in your reporting tells you about the sale you were never in the running for. The demand spike is real. Whether you are even eligible to catch it is decided in your catalog, at the SKU level, long before the shopper ever asks.
This year's event moved from July to late June. Adobe notes that shift helps make Q2 the first-ever $300 billion online quarter outside the holidays ($301.4 billion, up 11.9% year over year), with June spend alone up around 23%. It also drags back-to-school demand forward into June, well ahead of the usual window.
I do not want to overstate one date change, but step back and the pattern is bigger than Amazon's scheduling. Major promotional demand is decoupling from the seasons that used to define it. Holiday peak, summer peak, back-to-school arriving early, and increasingly, demand wherever a retailer decides to manufacture it. The calendar is becoming more of a suggestion every year.
For commerce teams, that subtly rewrites the job. The old rhythm was to spin up catalog readiness around a few tentpole moments, push the enriched data and the deal content ahead of the spike, then exhale. AI shopping agents do not honor that rhythm even slightly. They answer whatever a shopper asks, whenever they ask it, with zero awareness that it is supposedly "not back-to-school season yet." A catalog that is only accurate during the run-up to a known event is exposed the rest of the year, which is now most of the year.
So readiness stops being a seasonal project and becomes a standing posture. Your product data has to be interpretable, complete, and current all the time, because the next spike is no longer on a schedule you control.
That is the part I find genuinely exciting, and I realize that probably says something about me. The people writing the carousels are optimizing for the last decade of retail. I would rather spend my time on the version that is actually arriving, lunch boxes and all.
All figures are from Adobe's U.S. Prime Day 2026 forecast (June 23 to 26, 2026) and Adobe's Q1 2026 AI Traffic Report. If you want the ongoing version of this conversation, Scot writes about agentic commerce every week over at Retailgentic (he covered Adobe's 33x Prime Day GenAI forecast last year, too).