Tracking Agentic Commerce Through Black Friday and Cyber Monday: What Our Traffic Data Shows

The five-day period from Thanksgiving through Cyber Monday has long been considered a cornerstone of the holiday shopping season. But this year, it’s taken on a new importance as the first true trial-by-fire for agentic commerce. AI platforms and major retailers alike have embraced agentic commerce in Q4, partnering on milestone releases like ChatGPT’s Instant Checkout and integration with Target and Walmart, as well as PayPal and Perplexity’s Instant Buy program.

But the true pace of consumer adoption for agentic commerce tools can be hard to pin down. Our recent survey shows that almost 65% of consumers say that they’re interested in engaging with AI tools such as ChatGPT to do their holiday shopping this year. But has that played out?

At HUMAN Security, we verify more than 20 trillion digital interactions each week, giving us unique visibility into agentic traffic and market trends. We can see which agents are visiting sites and applications, the paths they travel, and the actions they take. We can also see how quickly agentic traffic is accelerating. We’ve also built AgenticTrust, the first solution purpose-built to provide visibility and governance over AI agent activity.

Here is what our telemetry shows:

• Agentic traffic grew 6,900% year-over-year, up sharply from the 1,300% growth observed earlier this fall.
• Month-over-month volumes have risen 72.4% on average since July, more than doubling again from September to November.
• During the 5-day period from Black Friday to Cyber Monday, agent traffic across all clients rose ~28%, while traffic targeting e-commerce sites surged 144.7% compared to the previous 5 days.
• 83.3% of AI agent activity during the Black Friday/Cyber Monday sales period focused on product and search pages, a clear shift toward high-intent shopping.
• Payment and checkout attempts more than doubled, moving from <1% of traffic in October to >2% during the holiday period.

Let’s dive a bit deeper into the numbers. 

Note: The telemetry presented in this post describes trends across HUMAN-verified interactions, not an absolute representation of all AI agent activity or the market share of all agents.

Holiday shopping trends show clear growth for agentic commerce 

Agent deployment rose meaningfully across the entire market heading into Cyber Week. Across all clients, agent traffic increased by more than 28% from the pre–Black Friday baseline (Nov 21–26) to it’s peak on Nov 28, and sustained higher traffic levels throughout the shopping period before returning to normal volume on Tuesday, December 3rd. This reflects a broad, industry-wide ramp-up in the use of autonomous AI tools during the holiday period.

But traffic to retail and e-commerce sites paints an even clearer indication of commercial intent. Agent traffic targeting e-commerce clients exploded on Black Friday. The daily traffic volume directed exclusively at shopping sites increased by 144.70% from the pre-BFCM baseline to the peak day (Nov 29), and sustained highly elevated levels throughout the sales period. 

How AI agents’ actions tell the story of agentic adoption

If rising traffic volumes show that agents were deployed more widely during Cyber 5, their behavior shows why. The holiday weekend produced a sharp shift back toward commerce-focused activity: more than 83% of AI agent traffic went to product and search during Black Friday weekend, up from 51% in the previous month. This is a big deal for online retailers, as agent-driven recommendations convert at 4.4x the rate of traditional search. All in all, Adobe reports that retailers brought in  $44.2 billion online during the Black Friday/Cyber Monday weekend.

To understand how meaningful this shift is, it helps to look at how agent behavior evolved throughout the year. You can click through the tabs of the pie chart above to compare how AI agents have been used throughout H2 of 2025.

Back in August, the vast majority of AI agent traffic went to product and search pages (86.6%), indicating that agents were primarily conducting research, comparing items, and scanning product options before narrowing down to specific choices. Checkout activity was rare, and account interactions such as login, profile, or order history pages made up only a small fraction of traffic. 

In September and October, the behavioral mix changed. Product and search pages still represented a meaningful portion of activity throughout the fall (56.2% in September and 51.1% in October), but a growing share of traffic moved to login and account pages as well (40.4% in September and 39.5% in October). This change signaled a new usage pattern: consumers were increasingly relying on agents for actions behind login and productivity, not just commerce. In other words, agentic browsing started to look less like a shopping assistant and more like a productivity layer integrated into everyday digital routines.

During the holiday sales period, the pattern flipped again. We saw a dramatic move towards product pages as agents actively assisted human holiday shoppers. Product and search pages comprised 83.3% of AI agent traffic. AI agents were primarily sorting through inventory and providing information to users once again, but this time will clear purchase intent. 

Checkout behavior also changed meaningfully. Payment and checkout attempts more than doubled during this period, rising from less than 1% of traffic in October to more than 2% in late November. The absolute share remains small, reflecting that many consumers are still hesitant to delegate final purchase authority to AI. But the acceleration shows how quickly agents move toward conversion behavior when user intent is time-sensitive and deal-driven. This pattern indicates that agent-assisted checkout is likely to grow as consumer comfort increases.

Agentic traffic is growing exponentially in 2025

To understand why agentic traffic surged so dramatically during Cyber 5, it helps to examine the trajectory leading into the holiday season. The underlying growth curve has been steep all year. 

In September, HUMAN research revealed that agentic traffic grew more than 1300% in the first eight months of 2025. That trend has only accelerated. 

We now observe a 6,900% growth in AI agent traffic YoY, driven by the release of new agentic browsers and the introduction of agentic commerce partnerships that channel real consumer shopping intent through automated browsing tools.

Month-over-month data shows the same pattern of rapid compounding. Total agentic traffic more than tripled from July to September and then more than doubled again from September to November. Across this period, agent volumes have grown by an average of 72.4% month-over-month, evidence that adoption is neither linear nor episodic, but exponential. 

The changing composition of agentic traffic

The mix of agentic traffic has shifted month over month as new browsers enter the market and platforms push deeper commerce features.

In July and August, ChatGPT Agent dominated activity, accounting for 88% of agentic traffic in July and 81% in August. Other players were small and inconsistent, and the market functioned largely as a single-entrant environment.

By September, that changed. Perplexity Comet captured 50.95% of all agent traffic. Comet’s early access programs and free account distribution accelerated adoption and quickly reshaped the traffic mix.

October continued the shift. Comet reached 69.7% and ChatGPT Agent fell to 25.29%. Browserbase and Browseruse also began to appear more consistently as developers experimented with automation-focused tooling.

In November, Comet expanded its lead to 74.38 percent. ChatGPT Agent declined to 15.94 percent. Several new entrants, including ChatGPT Atlas, OpenAI’s agentic browser, appeared with small but measurable traffic levels. These tools are too new to influence market share, but their arrival signals the next stage of competition as more platforms add agentic browsing and commerce capabilities.

A small caveat

Agent traffic is still a small share of overall AI-driven activity. HUMAN processes trillions of interactions each week, but agentic traffic currently appears in the millions per month. Traditional scrapers and crawlers, including AI scrapers such as ChatGPT-User and ChatGPT-Bot, operate at far higher volumes. This scale difference can make agents easy to overlook in aggregate analysis, but their growth rate and economic impact make them essential to monitor.

It is also important to distinguish between types of AI-driven commerce activity. Not all shopping interactions use full autonomous agents. AI chat interfaces often rely on a mix of retrieval-based scrapers and lightweight automation to surface products or prices. The research in this post focuses specifically on autonomous agents and agentic browsers. In future publications, we’ll compare these patterns with RAG-driven tools that support real-time product research and comparison, and give an even broader view of the agentic commerce landscape, but currently, the smaller scale of true agentic interactions would make such a comparison meaningless. 

On the potential risks of agentic commerce

It’s clear that agentic commerce is already shaping how consumers research and buy. Businesses that adapt early can gain real commercial advantage, but embracing agentic traffic without visibility or governance can leave you vulnerable to fraud and abuse. 

Here are some examples of the types of risk exposure that could arise.

• Shifts in traffic and demand signals. Agentic systems can compress complex browsing and comparison steps into rapid sequences that resemble normal user behavior. When this automated activity is not recognized as such, it can distort analytics, strain pricing and inventory services, and skew experiments or personalization models.
• Actions taken inside user sessions. Agents now operate inside authenticated environments. They can modify accounts, adjust carts, or submit forms based on broad user intent rather than explicit confirmation. When outcomes diverge from what the customer expected, the business usually absorbs the blame.
• Blended traffic that evades controls. Most agentic clients look identical to mainstream browsers. Traditional controls like user-agent filtering or IP heuristics offer little discrimination. Without explicit classification, businesses cannot tell how much of their critical workflow volume is being driven by automated intermediaries.
• Expanded security and operational exposure. Agents interpret on-page content and interface cues as potential instructions. Misconfigurations, permissive endpoints, or adversarial patterns can trigger unintended navigation or excessive interactions with sensitive flows. Automated traversal amplifies weaknesses that once affected only human users.

How to take advantage of agentic commerce

Is the answer to block all AI agents? That might make brands more secure, but also harms their bottom line. If agents can’t discover products on your website, they won’t recommend them. This may cause consumers to shop with competitors whose product catalogue agents can access. 

Blocking agents also frustrates consumers who want to use these tools to shop. In order to keep their competitive edge and preserve consumer loyalty, retailers must embrace agentic commerce. But they must do it safely. This starts with trust.

The first step is to verify that AI agents are genuinely who they claim to be (e.g., that they’re not spoofed or impersonated) and assess their trustworthiness. Then comes full visibility into agentic traffic, including behavior, actions, and intent. Finally, set clear rules about what agents are allowed to do and what should be blocked—for example, letting verified agents browse and discover products, while preventing them from creating new accounts. 

The result is confidence that every agent interaction is both authentic and aligned with the business’s security policies and risk standards.

The holiday shopping season is just the beginning

Looking back on Cyber 5, one thing is clear: AI agents are becoming a real shopping tool. HUMAN’s data shows that consumer adoption of AI agents has only been increasing month over month, and the post-Thanksgiving leap suggests that this holiday shopping season will be one for the books. But it won’t end there.

The merchants who will come out on top aren’t just tracking agentic commerce trends; they’re monitoring agent behavior in real-time and dynamically governing it. That means verifying which agents are visiting your site, understanding their intent across the customer journey, and enforcing granular controls over what they’re allowed to do.

Safely enable agentic commerce with HUMAN

HUMAN AgenticTrust provides session-level visibility into which AI agents are browsing your catalog, attempting logins, or reaching checkout. It allows merchants to understand agent behavior patterns, evaluate trustworthiness, and identify whether activity was user-prompted or autonomous.

The solution cryptographically verifies trusted agents and enforces action-based controls to govern what agents can and cannot do. Customers can set granular rules for each agent, allowing one to freely browse but not create an account or check out, while choosing a different set of guardrails for another agent. With complete visibility and control of AI agent actions, retailers can securely allow AI and agents to discover products and make purchases on their business terms. This level of granularity is necessary for brands to safely generate revenue from agentic commerce.