# The Agentic Internet Needs Labelled Influence, Not Fewer Machines *Bots now outnumber humans on Cloudflare's network and agents are starting to shop. The real risk is not fewer humans online. It is influence over agents that nobody can see or price.* Published: 2026-10-26 | Read time: 5 min read | Author: Smalk AI Research | Source: https://www.smalk.ai/blog/agentic-internet-labelled-influence-not-fewer-machines --- The agentic internet does not end the human web; it ends the human reader as the unit that advertising and publishing were priced on. The Arab Weekly's October 26, 2026 essay asks whether this is the end of the human internet. The sharper question it buries in its own argument is who decides what the agent reads, compares and recommends, and whether anyone can see that influence being bought. - Cloudflare reported on September 30, 2026 that daily AI-agent requests on its network grew more than 1,700% in a year and that over half of its traffic is now non-human. - The browsing web monetized human attention with labelled ads next to ranked links; the agentic internet routes attention through software that reads sources and presents one recommendation. - Influence over that recommendation is already being bought, through paid content, private licensing and data feeds, and none of it is labelled to the person who receives the answer. - A disclosed, priced ad layer inside the content agents read is more accountable than unlabelled influence, and it is the only version that also pays the publishers whose pages build the answer. ## What the agentic internet is The agentic internet is a web whose fastest-growing audience is software acting on behalf of people: agents that search, read, compare, fill forms and increasingly pay. Cloudflare described it in August 2026 as readable, discoverable, callable and payable, and in September 2026 called agents the internet's second audience. That audience does not see page design, does not scroll and does not click banners. It reads text, weighs sources and hands a human a conclusion. ## The browsing web vs the agentic internet The browsing web sold human attention in a visible place. A person typed a query, saw ten links and some labelled ads, chose one and read it. Influence was contested in public: ranking fights, ad auctions, paid inclusion scandals, and a rule that paid placement had to say so. The agentic internet moves every one of those steps upstream, into the sources an agent retrieves and the weight it gives them. The human sees the conclusion, not the contest. What carries over is the commercial incentive to shape the result. What disappears is the label that told you when shaping had been paid for. ## What The Arab Weekly reported, and what we verified The Arab Weekly argues that AI agents are becoming the primary intermediary between humans and the web, that publishers lose the visit when an agent reads for the user, and that control over information is shifting from search engines and social platforms to agents. It cites Cloudflare's 1,700% agent growth and majority non-human traffic, Chartbeat data reported by Axios on falling search referrals, and Visa's view that agent-initiated transactions could become a major category. Cloudflare's September 30, 2026 network report confirms both traffic figures. Chartbeat's 2026 Publisher Playbook puts Google Search referrals to publishers down 40.2% between July 2025 and July 2026, with AI chatbots sending 0.01% of publisher traffic. One clarification: the September 19, 2026 Axios story on collapsing news traffic relies on Similarweb, which shows top US news sites falling from roughly 68 million visits in November 2024 to 47.6 million in August 2026, rather than on Chartbeat. Visa reported live agent-initiated purchases with more than 30 European issuers on July 2, 2026 and expects millions of consumers to buy through agents by the 2026 holiday season. ## Why unlabelled influence is the real risk, not fewer humans ### The strongest case that the human internet is fine Cloudflare CEO Matthew Prince told NBC News in June 2026 that agent capability disproves the dead internet theory rather than confirming it: machines browse so humans can act. Chartbeat's 2026 data adds that overall publisher pageviews held near 31 billion a month even as search referrals fell, carried by loyal readers, email and apps. On this reading, the agent is a tool, the human still decides, and the open web adjusts as it did to mobile and social. ### Why the control question is a disclosure question That defence is correct about humans and wrong about accountability. The human still decides, but from a shortlist the agent assembled from sources it chose, and those choices are already for sale. Publishers sell GEO-optimized branded content that aims at the retrieval layer. Google pays selected publishers for content that shapes AI answers on terms set in private. Retail feeds decide which products an agent can even compare. None of that reaches the user as a label. The browsing web survived twenty-five years of paid influence because the influence was marked. The agentic internet is repeating the economics without the marking, and The Arab Weekly's worry that users may not know why an agent chose one product is the precise symptom. > The last word, payable, is the one that determines whether the open web can fund itself. > > — Cloudflare, The Internet has a second audience, 30 September 2026 Payable, in Cloudflare's design, means an agent pays to access a page. That funds access. It does not make influence visible, and it does not pay the publisher for being the source that won the recommendation. Labelled influence, bought in an open market and disclosed in the content the agent reads, does both. ## What this means for brands and for publishers ### For CMOs, media buyers and agencies: buy influence you can audit The budget line that matters is influence over the agent's shortlist, not the ad beside the answer. Buy it where it can be labelled, logged and shown to a regulator, because the EU AI Act's Article 50 transparency duties applied from August 2, 2026 and the IAB published an AI transparency and disclosure framework the same month. Audit how much of your consideration set is now assembled by agents, and treat undisclosed paths into that set as liability, not as a growth hack. ### For publishers: provenance is the product the agent will pay for The Arab Weekly is right that provenance becomes as valuable as content when machines summarize machines. A publisher whose pages carry verified identity, dated reporting and disclosed commercial messages is the source an accountable agent prefers and the source a brand can safely buy presence on. Keep verified agents on cited pages, mark what is paid, and sell that marked presence to brand demand rather than renting the archive to a single platform. ## Three signals that disclosure will decide the agentic internet First, identity is becoming infrastructure: Cloudflare now verifies more than 500 billion signed bot requests a week through Web Bot Auth, which gives publishers proof of which agent read which page. Second, regulators are moving from AI-generated content to AI-mediated recommendation; the FTC's Operation AI Comply has produced more than a dozen actions and guidance on sponsored placement inside agents is the obvious next step. Third, agentic checkout is about to produce receipts: once Visa's holiday-season purchases flow through agents, the line between sponsored and inferred picks will be a consumer-protection question, not a marketing one. ## Conclusion Hold on to this: the agentic internet does not threaten humans; it threatens the visibility of influence, because the contest over what gets recommended has moved upstream into the sources agents read, where nothing is labelled and nobody is paid. Smalk AI takes the position that the answer is a market, not a wall: Generative Engine Advertising places disclosed native ads for AI agents inside the publisher content those agents read and cite, so brands buy presence where recommendations form, publishers earn media revenue for being the source, and the influence is visible to the human who receives the answer. What to watch next is the first regulatory ruling on sponsored placement inside an agent's recommendation, because that is the moment unlabelled influence becomes a liability and labelled influence becomes the category. ## FAQ ### What is the agentic internet? The agentic internet is a web whose fastest-growing audience is software acting for people: AI agents that read, compare, call services and pay on a user's behalf. Cloudflare describes its requirements as readable, discoverable, callable and payable. For brands and publishers, the shift means influence happens in the sources an agent reads rather than on a page a human views. ### Is the human internet really ending? No. Humans still create, communicate and make final decisions online. What is ending is the human reader as the unit advertising and publishing were priced on. Cloudflare reported in September 2026 that more than half of traffic on its network is non-human, so the economics now have to account for a machine audience as well as a human one. ### How is advertising to AI agents different from advertising beside AI answers? Advertising beside an AI answer buys a labelled slot next to the result, much like a search ad. Advertising to AI agents places a disclosed brand message inside the publisher content agents read, so the brand is present where the shortlist forms and the cited publisher is paid. The second model reaches the recommendation; the first reaches only the screen. ### Should publishers block AI agents to protect human traffic? Blocking agents protects a shrinking asset. Chartbeat's 2026 Publisher Playbook shows Google Search referrals to publishers fell 40.2% in a year while AI chatbots sent 0.01% of traffic, so the human referral is not coming back. The durable position is to be a verified, cited source and to monetize agent reads directly rather than to lock them out. ### Why does disclosure matter more in an agent-driven internet? Because the user sees a conclusion rather than a results page, any paid influence on that conclusion is invisible unless it is deliberately marked. The EU AI Act's transparency obligations applied from August 2, 2026 and the IAB issued an AI disclosure framework the same month. Labelled influence is the only form regulators, brands and users can all audit.