Ari Abelson is Co-Founder of OpenOrigins, building provenance infrastructure for AI, media and the internet.
getty
The internet is being reshaped at a lightning pace. Already,
51% of internet traffic is bots, AI agent traffic is growing 8 times faster than human traffic, and there is an estimated
$40 billion of GenAI fraud losseshitting the U.S. by the end of 2027.
In other words, the internet is much less human and much more dangerous to navigate than ever before.
But the biggest problem is one hidden just underneath the surface: No one knows how to tell human traffic and AI traffic apart. We are all living on the same websites, all clicking the same buttons and all sharing content on the same feeds.
We have lost the ability to tell humans from AI, and the consequences will be dire if we don’t solve this problem soon.
In a world of humans, the internet’s intelligent user base was capped proportionally to the population of people online. With AI, the number of agents performing “human-like” actions is essentially infinite, dependent only on global compute.
It isn’t unreasonable to imagine trillions of agents navigating the internet in the coming decade. They will be collating information on the stock markets and deploying capital autonomously, picking out patterns, fabrics and factories for companies to use and planning, finding and booking our travel. The potential online actions are limitless.
On one hand, this future allows humans more time to do the things we love (which, for a few of us, is the administrative task of being alive). On the other hand, it’s an unbelievable security risk where agents socially engineer and coerce human beings through agent-to-agent fraud and manipulation.
And we are already seeing the consequence of this spiral.
Without the infrastructure to identify who is actually engaging with content, global ad ecosystems need to reevaluate the metrics they rely on to define success and determine what truly merits ad spend. For example, a recent study found that 8.5% of digital ad traffic is invalid, resulting in $63 billion of wasted spend.
At the same time, not all nonhuman traffic is a problem. In fact, recent AI traffic appears to be commercially promising. AI-driven traffic to U.S. retail sites rose 393% year over year in Q1 2026, and agentic shoppers now convert 42% better and spend 37% more per visit than humans.
The real issue is that advertisers cannot yet distinguish between human and agent audiences, and importantly, we can’t yet tell which agents are working for authentic human buyers. This makes it difficult to allocate spend intelligently. Until ad tech can verify not just whether a visit came from a person or an agent, but also the intent behind that interaction, billions will continue to be wasted on traffic that is either nonexistent or misaligned with real demand.
The most quietly devastating consequences are personal. As humans become the minority on the internet, the exponential rise of AI and agentic traffic has created a perfect storm for online intimacy. A trial conducted by MIT Media Lab and OpenAI found higher daily chatbot use correlated with greater loneliness, emotional dependence and reduced real-world socializing. The FTC reported $2.1 billion in scam losses originating on social media in 2025, eight times the 2020 total.
There are two problems compounded together here. First, we need to bifurcate the internet so humans can socialize with humans, knowingly. Ensuring people aren’t defrauded on a mass scale.
The other is that we need auditability of AI agents and companions. It’s not enough to put in safety rails; we need to regulate accountability deeply.
The most important consequence of generative AI has been its impact on trust. Trust depends on knowing what you are looking at, who you are talking to and if the intent they have aligns with your own. As AI-generated content saturates the web, our trust in what we see is collapsing.
We are entering a perfect storm. Institutions are less trustworthy:
Trust in national newsdropped 11 percentage points from March to September of 2025 alone. Seventy-four percent of people have identified social media as the least trusted environment.
Right now, AI slop has taken over our shared social spaces.
Over 50% of the internetis now estimated to be AI content, and
53% of Americansaren’t confident they can tell the difference between humans and AI.
So we can’t trust institutions, and we can’t trust user-generated content. This problem will only compound as all these tools get more sophisticated.
So what’s the answer?
AI is not the problem—it’s an amoral tool: Autonomous agents, synthetic content and bot-driven systems can be genuinely useful.
The problem is the absence of infrastructure to tell the difference. A world where AI and human content coexist is fine, as long as you can tell which is which.
The only durable solution is new infrastructure that can prove what content, engagements and relationships are of human origin. We need to be able to hash, catalog and prove every single transaction, piece of content and agent online. We need to build digital passports for every online actor (agent or human), every online piece of content (human-created) and every handshake.
The internet of old is out. We now need proof of every single datapoint that runs through it.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。