The Human-First Internet Is Over. Is Marketing Ready?

- By Michael Cucchi, CMO, Hydrolix

I recently attended a meeting in Q1 when somebody casually said that non-human traffic has exceeded 50% of all Internet traffic. No reaction from anybody in the meeting room. That was the moment that stuck in my mind, because virtually the entire marketing function is premised upon the assumption that there is a person at the receiving end of communications. The assumption that there is a person at the receiving end of communications is over.

AI bots are no longer just a threat to the security posture of a firm. They are a threat to the overall business model of a firm (and in many cases a boon to a firm depending upon how a firm chooses to react). Most marketing organizations are not architected to react to the change posed by AI bots.

From a cost perspective, according to Imperva's 2025 Bad Bot Report, bad bots accounted for 37% of all Internet traffic. Statista estimates that digital ad fraud will increase from $88 billion in 2023 to $172 billion by 2028. This means that bad bots are paying for digital ads and generating false impressions while also consuming content that was designed (and paid for) to be viewed by humans. This is a fundamental flaw in how we measure (roi), protect (digital spend), and optimize (ad performance) our digital investments.

The tools we have developed to combat this type of traffic have not kept pace with the increasingly sophisticated nature of modern AI powered bots. As such, captcha technology and human verification challenges are archaic. Modern AI powered bots are capable of mimicking human behavior so effectively that they can bypass most of the defense mechanisms that we have put into place. Our data at Hydrolix indicates that most web application firewalls (WAFs) fail to recognize more than half of bot traffic -- an alarming number for every marketer and security professional in attendance.

Silos have further complicated matters. Each of the three primary groups involved (security, operations, and marketing) have vastly different opinions regarding what constitutes "good" vs. "bad" bots. These groups use different tools; they operate on different schedules; and ultimately, they may arrive at opposing conclusions as to whether certain types of traffic are good or bad. Unfortunately, none of these groups share a single, cohesive view of the data available to help each group agree on whether certain types of traffic constitute "bots."

The misstep resulting from each group's incorrect determination cuts both ways. Misclassify too aggressively and you will eliminate true-value-driving bots (search crawlers, contextual verification tools, AI-indexing engines that determine whether your brand appears within LLM-generated answers). For example, one organization blocked valid global web crawlers and subsequently lost approximately $6 million dollars. Why? Because the crawlers did not appear in places they needed to appear.

Misclassifying bots incorrectly in either direction is costly, not to mention we are living in a next-generation, agent-based economy. My next big "aha!"  moment came when our product team approached me and asked how marketing could evolve if we began shipping new capabilities on a daily basis. Their AI-fueled development speed was so rapid that we had to inquire how we could evolve and accelerate every other aspect of the business to match its pace. And it was during that discussion that I understood all aspects of work and, indeed, all software-based interactions will soon transition into an agent-to-agent economy.

The AI-bot challenge facing marketers today is simply the precursor to this eventual reality. Several of the world's largest e-commerce organizations have already created agent-to-agent marketplaces. No humans in the middle. These exchanges occur at speeds previously unheard-of in terms of transactional frequency. When humans browse websites, it typically takes seconds for pages to load. However, when an agent makes a purchasing decision, it expects millisecond response times.

Financial markets were ahead of us. Initially, only humans traded stocks. Then larger institutional players accumulated capital and placed larger bets. Eventually, they pioneered machine learning algorithms to quickly decide what to buy/sell and when. Ultimately, they implemented AI systems that listened to earnings calls, inferred subsequent actions to take and took those actions in near-real time. Today, financial markets are experiencing exponential growth in transaction throughput. The exact same trajectory is unfolding in e-commerce and martech stacks -- and likely far sooner than most anticipate.

Impact Upon Marketing

The implications for marketing are straightforward: everything you currently understand about calculating roi, evaluating ad effectiveness and optimizing digital expenditure must begin moving towards real-time measurement. Hourly reporting cycles are already antiquated; they will be rendered irrelevant in an agent-to-agent world.

Where to begin…to be blunt, I'm still unsure how best to solve the dual-audience content conundrum. However, those marketers who are asking this very question are positioned better than those who are not.

Here is where I recommend we start:

1. Behavioral detection must replace signature detection. Modern agentic bots do not respect robot.txt files nor do they broadcast their presence through user-agent strings. Instead, they interact dynamically based upon their current state and request additional data sets in real-time. Therefore, detecting them via signature detection alone is ineffective. Behavioral analytics are necessary -- i.e., determining what an agentic bot does rather than what it looks like. Currently, most solutions on the market lack sufficient behavioral intelligence to accurately identify next-generation agentic bots.
2. Two audiences require different content strategies. When humans come to your website, they need compelling (emotional) content that promotes your brand. Conversely, when AI systems come to your website to index information or perform tasks, they require structured, actionable content that allows them to complete their objectives quickly. These represent fundamentally different forms of content -- treating them similarly will result in failing both audiences simultaneously. While optimizing for LLMs so your brand surfaces within AI-generated answers represents a current challenge, not merely a future consideration, optimizing for LLMs will be essential regardless.
3. Real-time data fabric is required. Traditional architectures cannot support large-scale bot traffic. Thus, most firms will either impede the performance of their AI workflows or generate invoices that will harm their businesses. Enter the real-time data fabric -- a layer built atop traditional architectures that collects and retains data from multiple sources (CDNs, WAFs, etc.) and delivers insights to inform AI workflows in real-time. This layer enables firms to quickly provide AI with the specific data it requires to perform its intended functions; once an AI receives the requested data set(s), firms can then monitor whether the interaction produced measurable outcomes.

Why are financial markets relevant?

While there is no identical parallel between financial markets and marketing/e-commerce organizations, they offer perhaps the most similar glimpse into what lies ahead. Firms in financial markets transitioned from human traders using intuition, to algorithmic systems relying on machine learning, to AI systems processing large volumes of data and producing actionable decisions in near-realtime. Similarly, firms transitioning from HumanFirst worlds to agentic economies will experience an analogous trajectory — however much faster.

Implications for marketers

For marketing professionals seeking success in the agentic era, accepting this inevitable trajectory early-on; auditing their bot/agentic strategies transparently; and expanding optimization efforts beyond  human-first worlds will prove essential.

Next steps? Learn your tools! Claude, ChatGPT, agentic platforms are not optional for future generations of marketing leadership; they represent your new ecosystem. Pressure vendors: directly ask your vendors: how does your tool integrate with agentic systems? If they cannot articulate clear responses -- that represents your response.