Tomorrow is November 1. By Adobe’s count, that is the moment the largest, most instrumented, and least verified shopping season in history opens for business.

The forecasts are in, and they agree on the shape of the demand even where they differ on scope. Adobe Analytics, working from more than one trillion visits to US retail sites, expects $275.1 billion in US online sales between November 1 and December 31, up 6.7 percent year over year. The National Retail Federation projects $1.013 trillion in total US holiday retail sales, up 4.2 percent, on its Census-based definition. Deloitte brackets the online piece at $316.1 to $318.9 billion across its November-to-January window. Salesforce, rather than publishing a sales total, made the boldest structural call of the season back in July: 20 percent of holiday e-commerce traffic will come from AI agents, and one in three e-commerce sites will have its own shopping agent live by Cyber Week.

Four forecasts, four methodologies, one shared blind spot. Every one of them measures traffic, conversion, and spend. Not one of them measures whether the product evidence underneath it all, the reviews and ratings that shortlists are built on, is real. This holiday season is the first in which that omission has teeth, because this is the first season in which a measurable share of the buying will be done by machines that cannot tell the difference.

The record everybody agrees on

Start with the demand picture, because it is genuinely historic. Per Adobe’s September 28 forecast:

  • Cyber Week (Thanksgiving through Cyber Monday) will drive $47.5 billion online, up 7.4 percent, or 17.3 percent of the entire season in five days.
  • Cyber Monday, November 30, is expected to cross $15.1 billion, up 6.2 percent, a new single-day record for US e-commerce.
  • Black Friday, November 27, will hit $12.9 billion, up 9.2 percent, growing faster than Cyber Monday.
  • Thanksgiving Day itself: $6.9 billion, up 8.5 percent, with a record 63 percent of that spend on mobile devices, the impulse channel if there ever was one.
  • The pre-season is no longer a warm-up. October alone should clear $95.8 billion, up 8 percent, with $9.9 billion across the October 6-7 Prime Day event. Adobe’s own post-mortem, tallied by Arktis from Digital Commerce 360 data, put the actual event at $9.86 billion, up 8.5 percent: the forecasts are already being validated in real time.
  • Electronics ($63.3 billion) and apparel ($51.3 billion) remain the tentpoles, with grocery the fastest-growing major category at $26.1 billion, up 10.3 percent.
  • Discounts peak at 30 percent off list during Cyber Week in electronics, 29 percent in toys, 23 percent in TVs, computers, and apparel.
  • Buy Now Pay Later will carry $21.3 billion of the season, including $1.09 billion on Cyber Monday alone.

Note that last item, because it is the pattern in miniature. The financing of holiday purchases is measured to the tenth of a billion. The authenticity of the products being financed is measured nowhere at all.

The overlay nobody had last year

What separates the 2026 season from every prior forecast cycle is the AI traffic column. Adobe now tracks when shoppers click through to retail sites from AI services, and its projection is stark: AI-referred traffic to US retail sites will rise 130 percent year over year across the season, with the steepest spike on Thanksgiving Day at 159 percent growth, 95 percent on Black Friday, and 88 percent on Cyber Monday.

Pause on the timing. The biggest AI-assisted surge lands on Thanksgiving, the highest-impulse, most mobile, least-deliberate shopping day of the year, when 63 percent of spend flows through a phone and the average session is measured in spare minutes between courses. That is precisely the context in which a shopper, or an agent, is most likely to accept a shortlist at face value.

FirstPageSage’s October agentic commerce study, built on a panel of 4,936 US shoppers, order data from 43 brands, and 18,400 shopping prompts run between July 13 and October 16, supplies the funnel underneath that traffic:

  • 61.4 percent of US online shoppers now use AI for product research, up 5.8 points since July.
  • 48.7 percent use it for price and feature comparison; 39.2 percent now read AI-generated review summaries, up 4.1 points.
  • 14.6 percent let an agent build a cart (up 55 percent in relative terms), and 6.8 percent approve an agent checkout (up 134 percent).
  • Fully autonomous purchase, no approval at all: 1.3 percent.

The study’s own summary of the approval economy is worth quoting directly:

“Shoppers overwhelmingly want a final say: agent checkouts with shopper approval outnumber fully autonomous purchases by more than five to one.”

Agent-completed orders rose from 0.27 percent to 0.68 percent of all US e-commerce orders between the second and third quarters, the largest quarterly jump in FirstPageSage’s dataset, with 71 percent of that increase traced to Meta Muse’s September 8 launch. Off a small base, that is the knee of an S-curve, and the holiday season is about to pour record volume through it. A separate structural finding deserves more attention than it has gotten: merchants inside an agent’s checkout partnership saw purchases complete at 1.4 to 3.8 times the rate of merchants outside it. Placement inside the agent is becoming the whole game. The question of what earns that placement, merit or money, is the one nobody is auditing.

The confidence trap

Buried near the end of Adobe’s release is a consumer survey that should unsettle anyone building agentic commerce, and it reads like a warning label for the entire season. Of 5,000 consumers surveyed, 77 percent of those who have used AI for online shopping report feeling more confident in their purchases, and 69 percent say they are less likely to return an item they bought with AI assistance.

Confidence is not verification. Nothing in the AI shopping stack checks whether the reviews behind a product rating are authentic; the models summarize what is there. And what is there is heavily contaminated. Fakespot-derived analyses have long estimated that roughly 30 to 42 percent of Amazon reviews are inauthentic, a figure Amazon disputes. Amazon’s own enforcement scale, 275 million suspected fake reviews blocked in a single year, is the clearest admission that the problem is industrial. Now layer on what the 2026 season adds: 39.2 percent of shoppers are having those same reviews pre-digested into confident-sounding summaries by models that treat every five-star review as equally credible. The AI does not just inherit the contamination. It launders it into fluency.

The FTC will be watching this season with penalty authority it did not have three years ago. Under the Consumer Reviews and Testimonials Rule, knowing violations now carry civil penalties of up to $53,088 per fake review, and the Commission signaled its seasonal posture last year by sending warning letters to ten companies during holiday shopping. But enforcement, by design, arrives after the purchases. It cannot inspect a shortlist at 11 p.m. on Cyber Monday.

Where the machine meets the manipulation

The holiday season concentrates three forces into the same four weeks: peak discounts, peak automation, and peak fraud.

The discount map tells you where agent attention will pool. Electronics peak at 30 percent off on Cyber Monday; toys hit 29 percent; TVs, apparel, appliances, and furniture post their best prices on Black Friday. Agents, whose objective functions overwhelmingly reward price and rating, will swarm exactly those categories on exactly those days.

Meanwhile the supply side of the scam economy has gone industrial ahead of the season. The Savi Scamwise Holiday Shopping Scam Report 2026, built on more than 250,000 shopping sites submitted since March, found that a brand-new fake store now appears roughly every three minutes, over 100,000 submitted sites were judged scams, 80 percent of scam stores lead with an impossible discount, and holiday-themed fake storefronts have multiplied fivefold since August. Seven in ten vanish within 90 days, roughly the window in which chargebacks surface.

Connect the mechanism. A price-optimizing agent that gets blocked or rerouted from major retailers, a dynamic this publication documented last week as Amazon, The RealReal, Tapestry, Kohl’s, Delta, and several major banks locked their doors to third-party agents, does not stop shopping. It goes where the prices look best. The fake store’s entire business model is being the best price an agent will find. Forter’s October data made the routing visible: screen-scraping agents, which roam the open web beyond the official commerce protocols, accounted for 70 percent of agentic orders by the week of October 12. The official rails are gated. The open web is not.

The column none of the forecasts contain

Return to the four forecasts and look for the evidence metric. Adobe models traffic, discounts, categories, BNPL, mobile share, even which gaming consoles will sell. The NRF models total retail and seasonal hiring, 325,000 to 375,000 workers. Deloitte models e-commerce brackets. Salesforce models agent traffic share and site-level adoption. In the combined hundreds of modeled variables across the season, there is not one that answers the question every one of these transactions silently depends on: is the review record of the product being bought authentic?

That is not a criticism of the forecasters. It is a measurement of the gap. The industry instruments demand beautifully, financing adequately, and evidence not at all. Into a $275.1 billion season, the evidence layer enters as the only structural input with no baseline, no forecast, and no owner among the platforms that monetize the purchases.

That is the gap GoBuy was built to fill, before the season rather than after it. Fake reviews are filtered out before any score is computed, authentic substantive reviews are weighted up, and every product gets a Smart Score from 0 to 100 based on review quality, not volume, so a listing with 40,000 reviews of which a third are fake cannot outrank one with 900 verified honest ones. Shoppers and agents get a curated top 7 instead of ten thousand sponsored results, because a shortlist you can verify is the only shortlist worth trusting at holiday velocity. Products holding a score of 80 or above across 90 days earn the GoBuy Verified badge, an integrity signal no marketplace grades itself on. Any AI agent can consult all of it live through the MCP server at gobuy.ai/api/mcp before committing a single dollar of someone’s holiday budget, and the Chrome extension puts the trust panel directly on the Amazon product page.

Tomorrow the season opens: $275.1 billion forecast, AI traffic up 130 percent, Cyber Monday set to break $15 billion in a day. Every dollar of it will ride on product evidence nobody has verified. The shoppers, and the agents, now have somewhere to check.

Before anything buys for you this season, check it first: gobuy.ai. Building a shopping agent? Wire in the trust layer before the volume arrives: gobuy.ai/agent-docs.