Something remarkable happened to the internet in 2025, and most organizations had no idea it was occurring. For the first time in the history of the web, automated traffic — driven overwhelmingly by AI systems, agents, and crawlers — overtook human-generated activity to become the dominant form of internet interaction. The data is not ambiguous. HUMAN Security’s 2026 State of AI Traffic and Cyberthreat Benchmark Report, released on March 26, 2026, and based on analysis of more than one quadrillion digital interactions, documents the inflection point precisely: AI bot traffic grew 187% from January to December 2025, while AI agent browser traffic specifically surged 7,851% year over year. Human traffic, by contrast, grew just 3.1%. Automated traffic is now growing eight times faster than human activity online.
The implications of this shift extend far beyond statistics. The internet was architecturally designed for human users — its performance characteristics, security models, content discovery mechanisms, and commerce infrastructure all assume a human being at the endpoint. When the majority of interactions are initiated by AI agents autonomously pursuing goals on behalf of users or organizations, every layer of that architecture faces stress it was not built to handle. Cloudflare CEO Matthew Prince, speaking at SXSW in Austin on March 19, 2026, put the trajectory in stark terms: AI bot traffic will exceed human traffic by 2027 — and the infrastructure of the web has no clear roadmap for absorbing that transition. Before the generative AI era, the internet was only about 20% bot traffic, with Google’s web crawler as its dominant non-human agent. That equilibrium is now broken, and it is accelerating.
This article examines what the AI traffic data shows, which industries are being most transformed, what the cybersecurity implications are, and how organizations should be thinking about defending and adapting to a web that is increasingly automated.
What Does It Mean That AI Bots Now Outnumber Humans Online?
For the first time in internet history, automated systems collectively generate more traffic than human beings. HUMAN Security’s benchmark data confirms that AI-driven traffic and other automation have crossed a critical threshold — the web is no longer predominantly human in its interactions, even though it remains predominantly designed for humans.
The distinction matters operationally because AI bots behave differently from human users at the infrastructure layer. A human researcher might visit five websites to answer a question. An AI agent fulfilling the same query will visit thousands. Prince’s estimate of 5,000 sites per query illustrates the traffic multiplication effect: as AI agents become the primary interface through which humans interact with the web, total traffic volumes will grow in ways that have no precedent in internet infrastructure planning.
| Traffic Type | 2024 YoY Growth | 2025 YoY Growth | Trend |
|---|---|---|---|
| Human traffic | ~3.5% | 3.10% | Stable, slow growth |
| Total automated traffic | ~8% | 23.51% | Accelerating |
| AI-driven traffic | ~60% | 187% | Exponential |
| AI agent browsers | ~200% | 7,851% | Near-vertical |
This table reflects a web that is being structurally transformed. The internet’s original architecture — TCP/IP, HTTP, DNS, TLS — was designed around the assumption of a human at the keyboard. None of those protocols anticipated a world in which the majority of requests come from autonomous systems operating at machine speed, machine scale, and machine autonomy.
flowchart LR
A[2022<br>Internet Traffic] --> B[80% Human<br>20% Bot]
C[2025<br>Internet Traffic] --> D[AI Agent<br>Traffic +7851%]
C --> E[AI Scraper<br>Traffic +187%]
C --> F[Human<br>Traffic +3.1%]
D --> G[2027 Projection<br>Bots exceed humans]
E --> GHow Fast Is AI Traffic Growing Compared to Human Activity?
Automated traffic grew 23.51% year over year in 2025, compared to 3.10% for human traffic — a ratio of roughly eight to one. But within the automation category, AI-specific traffic is growing at a rate that dwarfs overall automation figures.
AI-driven traffic — defined as traffic from LLM training crawlers, real-time AI scrapers, and autonomous AI agents — grew 187% in 2025 alone. AI agent browser traffic, which represents the most autonomous and commercially active form of AI-generated traffic, grew 7,851%. These are not gradual trends. They reflect the 2025 deployment wave of frontier AI models and the subsequent explosion in AI agent usage as enterprise deployments accelerated.
| Traffic Category | Annual Growth Rate | Commercial Activity Level | Primary Sources |
|---|---|---|---|
| Human users | 3.10% | Transactional | Individuals, enterprise employees |
| Traditional bots | ~8% | Indexing, monitoring | Search engines, uptime checkers |
| LLM training crawlers | ~60% | Non-transactional | OpenAI, Anthropic, Meta |
| Real-time AI scrapers | ~120% | Research, competitive | Multiple AI platforms |
| AI agent browsers | 7,851% | Fully transactional | Autonomous task completion |
The AI agent browser category is the most consequential. These are not crawlers reading content for training — they are agents transacting: logging in, browsing products, booking services, completing forms, and executing purchases autonomously. This is the traffic type that creates both the most commercial opportunity and the most acute security risk.
Which Industries Are Being Most Disrupted by AI Agent Traffic?
More than 95% of AI-driven internet traffic concentrates in three verticals, each of which shares a common characteristic: high-value, high-frequency transactional workflows that AI agents are naturally suited to automate.
Retail and e-commerce attracts the largest share of AI agent activity, driven by agents performing price comparison, inventory checking, procurement automation, and purchase execution on behalf of enterprise buyers and individual consumers. Streaming and media represents the second largest category, with AI agents consuming, summarizing, and redistributing content across downstream platforms. Travel and hospitality is the third concentration point, where agents book flights, hotels, and rental cars autonomously based on user or organizational preferences.
| Industry | AI Agent Activity Type | Commercial Impact | Security Risk Level |
|---|---|---|---|
| Retail and e-commerce | Price comparison, procurement, checkout | Revenue opportunity via AI buyers | High — fraud mirrors legitimate agent behavior |
| Streaming and media | Content consumption, summarization, redistribution | Licensing and traffic attribution complexity | Medium — scraping attacks on premium content |
| Travel and hospitality | Booking, price monitoring, itinerary management | Automated booking volume | High — credential stuffing on loyalty accounts |
| Financial services | Account monitoring, data aggregation | Emerging; growing rapidly | Very high — direct financial exposure |
| Healthcare and pharma | Research, formulary checks | Early stage | Medium — data sensitivity |
The three dominant verticals are notable not just for their AI traffic volumes but for the alignment between legitimate AI commerce and malicious automation. The same behavioral fingerprints — product discovery, login, checkout — describe both a procurement AI agent completing a legitimate purchase order and a fraud operation executing a carding attack. This alignment is the core operational challenge for enterprise security in 2026.
graph TD
A[AI Agent Internet Traffic<br>7851% Growth in 2025] --> B[Retail and<br>E-commerce 95%+ share]
A --> C[Streaming<br>and Media]
A --> D[Travel and<br>Hospitality]
B --> E[Legitimate<br>AI Commerce]
B --> F[Automated<br>Fraud Operations]
E --> G[0.5% behavioral<br>difference between E and F]
F --> G
G --> H[Enterprise Security<br>Classification Challenge]How Does the AI Traffic Surge Create New Cybersecurity Threats?
The security implications of AI-dominated internet traffic are not theoretical. HUMAN Security’s benchmark data quantifies a threat landscape that is deteriorating rapidly along every measured dimension.
The median percentage of global traffic attempting a scraping attack has nearly doubled since 2022, approaching 20% in 2025. Post-login account compromise attempts — the indicator of credential stuffing and account takeover attacks — more than quadrupled year over year, with HUMAN flagging an average of 402,000 such attempts per organization annually.
The most operationally difficult challenge is detection. Only 0.5 percentage points separate the rate of benign automation from the rate of malicious automation. Legitimate AI agents — shopping bots, research agents, procurement systems — use the same web paths, the same API endpoints, the same authentication flows as fraud operations. Traditional bot detection approaches based on traffic pattern analysis are becoming ineffective when both the legitimate and the malicious actor are operating AI-powered systems.
Three threat categories have emerged as the most pressing in 2026. First, account takeover at scale: AI-powered credential stuffing now tests thousands of login combinations per second across organizations with no rate limiting on post-login pathways. Second, AI-enabled web scraping: competitors and data brokers are deploying agents that systematically harvest proprietary product catalogs, pricing data, and customer-facing content at rates that overwhelm traditional anti-scraping measures. Third, prompt injection via web content: AI agents that browse arbitrary web content can be manipulated by malicious content embedded in pages they visit, redirecting their behavior in ways their operators do not intend.
Cloudflare’s response illustrates the scale of the problem: the company has blocked 416 billion AI bot requests since mid-2025 alone — a volume that reflects both the scale of AI traffic and the aggressive targeting of web infrastructure by automated systems.
What Did Cloudflare’s CEO Warn About the Future of the Web?
Matthew Prince’s SXSW remarks on March 19, 2026, represent the clearest articulation from a major infrastructure operator about what the AI traffic inflection point means structurally. His warning extends beyond traffic statistics to infrastructure architecture.
Prince’s core observation is that AI agents amplify traffic non-linearly. When a human user searches for information, they might visit five websites. When an AI agent fulfills the same information request — drawing on a large language model capable of browsing the web in real time — it may visit 5,000 websites. The multiplication factor is roughly 1,000x per query. At current AI adoption trajectories, this means internet traffic volumes could reach levels that existing infrastructure was never designed to handle, on a timeline of years rather than decades.
Prince’s proposed solution — sandboxed infrastructure for AI agents that can be spun up on demand and torn down after task completion — represents a fundamentally different internet architecture. Instead of AI agents freely browsing the open web as if they were human users, purpose-built agent execution environments would provide contained, metered, and authenticated access to web resources. This architecture would allow legitimate AI commerce to scale while providing meaningful control points for security enforcement.
The prediction also has significant implications for digital publishers, e-commerce operators, and any organization that relies on human web traffic for advertising revenue, audience measurement, or customer engagement. If the majority of traffic is generated by AI agents rather than human users, the metrics by which digital businesses have measured success for three decades become unreliable indicators of actual human audience.
How Should Enterprises Respond to an AI-Dominated Internet?
For organizations adapting to an internet where AI agents are a primary actor, the response has two dimensions: defensive, to address the security risks, and strategic, to capture the commercial opportunity.
On the defensive side, the most urgent investments are in traffic classification infrastructure capable of distinguishing between legitimate AI agents, benign crawlers, and malicious automation. This is technically difficult — as noted, only 0.5 percentage points separate benign from malicious automation — but several approaches have emerged as effective in early 2026 deployments. Zero-trust credential architectures, in which every authentication event is treated as potentially automated and verified with behavioral signals beyond the password, significantly reduce account takeover exposure. Short-lived access tokens scoped to minimum necessary permissions limit the damage radius if an agent’s credentials are compromised. Behavioral baseline monitoring, which establishes normal patterns for each category of automated traffic and alerts on deviation, provides early warning for both compromised legitimate agents and new malicious bot campaigns.
On the strategic side, forward-thinking organizations are beginning to optimize explicitly for AI agent consumers. The llms.txt standard — a web convention that provides AI systems with structured, machine-readable information about a website’s content — allows organizations to welcome legitimate AI commerce agents while directing them to intended pathways. Structured data endpoints, API-first content architectures, and explicit agent authentication programs represent the commercial opportunity embedded in the AI traffic wave: the organizations that make it easiest for legitimate AI agents to transact with them are positioned to capture a disproportionate share of AI-mediated commerce as the category grows.
The enterprises most exposed in the near term are those that have not yet distinguished between their web security posture for human users and their posture for automated systems. The internet is no longer a primarily human environment. Security strategies designed for human traffic patterns will fail against AI-scale threats at AI-scale volumes.
FAQ
What is AI bot traffic and how does it differ from traditional bot traffic? AI bot traffic refers to internet activity generated by AI-powered systems — including large language model crawlers, real-time AI scrapers, and autonomous AI agents — rather than by human users. Unlike older bots that simply followed fixed rules, AI bots reason, adapt, and transact on behalf of users or organizations, interacting with web platforms much like a human would but at machine speed and scale.
How much did AI agent traffic grow in 2025? According to HUMAN Security’s 2026 State of AI Traffic and Cyberthreat Benchmark Report, AI-driven traffic grew 187% from January to December 2025. More dramatically, traffic from autonomous AI agents and their browsers grew 7,851% year over year — a near-doubling of scale roughly every two months across the full year.
Which companies generate the most AI bot traffic on the internet? OpenAI dominates, generating approximately 69% of all AI bot traffic observed in 2025. Meta accounts for roughly 16%, driven by its AI training and inference infrastructure, while Anthropic accounts for approximately 11%. Together, these three companies represent the majority of AI-sourced traffic hitting websites and APIs globally.
How are AI bots creating new cybersecurity threats in 2026? AI bots and malicious automated traffic are nearly behaviorally identical — only 0.5 percentage points separate benign from malicious automation. This makes distinguishing legitimate AI commerce agents from fraud operations extremely difficult. Post-login account compromise attempts quadrupled year over year, and roughly 20% of global web traffic is now attempting a scraping attack at any given moment.
What did Cloudflare’s CEO predict about internet traffic at SXSW 2026? Cloudflare CEO Matthew Prince, speaking at SXSW in Austin in March 2026, predicted that AI bot traffic will exceed human traffic by 2027. He noted that before the generative AI era, the internet was roughly 20% bot traffic. AI agents amplify this because a single user query may cause an agent to visit 5,000 websites — 1,000 times more than a human researcher would.
Which industries are most affected by AI agent traffic growth? More than 95% of AI-driven traffic concentrates in three verticals: retail and e-commerce, streaming and media, and travel and hospitality. These sectors are being transformed simultaneously at the commerce layer — AI agents purchasing, booking, and transacting — and at the security layer, where the same agent traffic patterns are being exploited by fraud operators.
How should enterprises respond to an AI-dominated internet? Enterprises should implement AI-aware traffic classification that distinguishes between legitimate AI commerce agents, research crawlers, and malicious automation. Zero-trust credential architectures, short-lived access tokens, behavioral anomaly detection, and bot traffic telemetry are the primary defensive investments recommended for 2026. Businesses should also evaluate whether to formally welcome legitimate AI agents via llms.txt and structured data endpoints.
References
- HUMAN Security’s 2026 State of AI Traffic & Cyberthreat Benchmark Report — Primary source: full benchmark data on AI traffic growth and cybersecurity threat landscape
- AI and bots have officially taken over the internet, report finds — CNBC — CNBC coverage of HUMAN Security’s March 26, 2026 report
- Online bot traffic will exceed human traffic by 2027, Cloudflare CEO says — TechCrunch — Cloudflare CEO Matthew Prince at SXSW 2026
- HUMAN Security AI Traffic Growth Key Findings — Detailed methodology and industry breakdown
- AI bots now dominate the Internet, surpassing human traffic — Tech Startups — Independent analysis and enterprise implications