The Cyber 5 period of Thanksgiving through Cyber Monday once again delivered record online spending, even as online traffic softened across many retail sites. Shoppers were more selective this year — visiting fewer sites, comparing more across AI and traditional search, and converting at higher rates once they landed. With intent rising and traffic thinning, performance mattered more than ever.
The eCommerce Landscape for 2025 Cyber 5
Before the weekend, the National Retail Federation (NRF) reported that a record number of people — nearly 187 million — were planning to shop from Thanksgiving through Cyber Monday. It seems they did show up to go hunting for deals, as Adobe Analytics estimated $11.8 billion in online Black Friday spend. In total, consumers spent $44.2 billion during Cyber 5 2025, a 7.7% increase from last year.
But it wasn’t all good news for brands. Revenue growth masked softer fundamentals: online order volume and units per transaction both dropped year over year. Inflation, tariff pressures, and job instability pushed many consumers to tighten budgets and rely more heavily on buy now, pay later (BNPL), which hit a record $1.9 billion on Black Friday. Shoppers didn’t stop spending, but they became more selective in how and when they did it.
To break down exactly what happened, we looked at Yottaa data of millions of sessions on nearly 700 websites to see where performance held, where it lagged, and what we can learn about shopper behavior during the biggest digital week of the year.
Fewer Sessions, Stronger Intent
Traffic was down year over year, but the shoppers who showed up meant business. Yottaa data shows total sessions falling across every day of Cyber 5 2025 versus 2024:
- Thursday: –13.3%
- Friday: –12.6%
- Saturday: –16.6%
- Sunday: –15.0%
- Monday: –16.3%
But it doesn’t appear that the decline was a sign of weak demand. It was a shift toward more targeted, higher-intent shopping, with conversion rates rising throughout the weekend — especially on desktop, which peaked on Cyber Monday at nearly 6%.
Mobile Shopping Reigns Supreme
Although desktop wins on efficiency and conversion rate, mobile still drives the holiday weekend. Across every Cyber 5 day, mobile accounted for roughly 75–80% of all orders, making it the clear revenue engine of 2025.
Mobile conversion improved meaningfully year over year, rising from 3.3% in 2024 to 3.7% in 2025 — a 12% lift across the device that handles most online orders. Desktop remained the efficiency leader with a 4.93% conversion rate, but its performance softened slightly from 2024’s 5.4%, according to Yottaa’s data. The gap between devices narrowed, suggesting that shoppers are increasingly willing to complete purchases on mobile once the experience meets expectations.
AI Shopping Accelerates the Funnel
AI is estimated to have influenced more than $14 billion in global sales on Black Friday, with Adobe claiming that AI-driven traffic to retail sites increased a whopping 805% on Friday compared to last year. According to Yottaa’s data, AI discovery only makes up a very small percentage — less than 2% — of site traffic, but the behavioral patterns suggest early signs of rapid growth.
Even though AI referrals are a small percentage of total traffic today, their timing creates a new kind of peak-hour pattern. Traditional search peaks during afternoon promotions; AI peaks late at night, when fewer teams are watching dashboards and slower pages are less likely to be caught. From Thursday through Cyber Monday 2025, visits from AI platforms spiked every evening between 2:00 and 4:00 UTC, peaking on Black Friday evening. The pattern was remarkably stable: lower morning activity, steady afternoon buildup, and a sharp evening surge.
For digital operations teams, this adds a second performance window to monitor — one driven not by marketing events but by shopper behavior inside AI tools. As AI usage grows, overnight performance resilience will matter as much as daytime tuning.
Google Search, by contrast, remained the backbone of daytime discovery. Traffic peaked in the afternoon and early evening, aligned tightly with major promotional windows. Google also delivered the highest-converting traffic of the weekend, at 1.40%, and drove a broad mix of home, category, and landing-page visits.
Why AI Traffic Behaves Differently
The behavioral signature of AI shoppers is fundamentally different from traditional search — not because of who they are, but because of where AI drops them in the journey. AI referrals are overwhelmingly bottom-of-funnel, with 77% of visits landing directly on PDPs. That’s nearly 17 points higher than Google Search.
This shift to PDP-first entry compresses the buyer evaluation process. Shoppers bypass the exploratory layers of the site — navigation, search filters, merchandising modules — and land directly where specifications, pricing, reviews, and availability live. AI isn’t just moving shoppers deeper in the funnel; it’s skipping entire stages of the journey. That has implications for how brands structure PDP content, optimize pages, and protect performance when intent is highest.
Core Web Vitals Remained Steady
There was one notable difference between AI and traditional search shoppers: users arriving via AI referrals experience faster-loading pages than those coming from Google search throughout the entire Cyber 5 period. Across every day, AI referrals landed on pages with a p75 LCP between 1.4 and 1.6 seconds — roughly 20% faster than Google Search, which ran between 1.7 and 1.83 seconds.
CLS and INP told the same story. AI-referred traffic experienced more stable layouts (often 2–3× better CLS) and faster interactions (48–56ms). Meanwhile, Google-referred sessions consistently experienced 80–88ms. It’s possible these differences weren’t due to infrastructure strain but instead reflective of how AI and Google shoppers enter the site. AI referrals land primarily on PDPs, which tend to be leaner and less script-heavy, while Google continues to drive a broader mix of landing, home, and category pages.
Beyond channel-level differences, the overall performance environment remained remarkably stable when taking all sessions into account. Median LCP hovered between 2.42 and 2.49 seconds, median INP stayed in the 223–242ms range, and median CLS remained low and steady between 0.238–0.256, even as traffic, third-party load, and promotional activity fluctuated. These patterns are unusually flat for a period that typically pushes eCommerce infrastructure to its limits.
Even with the differences between AI and traditional search behavior, both segments and the ecosystem as a whole stayed well within “good” Core Web Vitals thresholds. There were no performance spikes, no late-night regressions, and no degradation during the heaviest shopping windows. For retailers on Yottaa, this indicates that site stability held even as shopper behavior shifted across channels, devices, and times of day.
Conclusion
This year’s Cyber 5 showed that the margin for error is shrinking. Shoppers visited fewer sites, used more tools to compare options, and converted quickly once the experience matched their expectations. With higher intent concentrated in fewer sessions, every slowdown, third-party delay, or layout shift becomes more expensive.
The fundamentals didn’t change — fast, stable digital experiences still win — but the context around them did. Discovery is splintering across AI and traditional search, buying windows are spreading into late-night hours, and mobile continues to dominate order volume.
Retailers that build performance resilience now will be the ones protecting revenue in 2026.
