Machine Media: The Death of the Open Web

The open web is dying as AI bot traffic on Cloudflare's network grew 187% in 2025 while human traffic grew only 3.1%, with agentic AI traffic surging 7,851% year over year. Global publisher Google traffic fell 33% in 2025, small publishers lost 60% of referral traffic in two years, and Cloudflare's CEO has stated bot traffic will exceed human traffic by 2027. The shift marks the arrival of Machine Media, where AI agents—not browsers or social feeds—become the primary interface between brands and customers, reducing websites to data sources and eliminating owned media as the destination.

The open web is dying. Not slowing. Not changing. Dying. For 25 years, every business model on the internet, from publishing to retail to SaaS to advertising, was built on a single assumption: that humans would visit your website. Owned media was the asset and the website was the storefront. Search, social, and email were the organic channels that filled it. That entire architecture is breaking down and owned media will no longer be the destination. Print. Broadcast. Web. Social. Machine. The next medium has already arrived, and the data is no longer subtle. AI bot traffic on Cloudflare’s network grew 187% in 2025 while human traffic grew 3.1% https://blog.cloudflare.com/radar-2025-year-in-review/ . Agentic AI traffic alone grew 7,851% year over year https://www.infoq.com/news/2025/12/cloudflare-2025-ai-bots/ . Cloudflare’s CEO has stated publicly that bot traffic will exceed human traffic by 2027 https://www.medianama.com/2025/12/223-user-driven-ai-bots-crawling-grows-15x-in-2025-cloudflare-report/ . Anthropic’s ClaudeBot now crawls roughly 24,000 pages for every single referral it sends back https://blog.cloudflare.com/ai-search-crawl-refer-ratio-on-radar/ . OpenAI’s GPTBot sits at 1,276 to 1 https://blog.cloudflare.com/crawlers-click-ai-bots-training/ . Global publisher Google traffic fell 33% in 2025 https://pressgazette.co.uk/media-audience-and-business-data/google-traffic-down-2025-trends-report-2026/ . Small publishers lost 60% https://www.searchenginejournal.com/search-referral-traffic-down-60-for-small-publishers-data-shows/569959/ of their referral traffic in two years. Some lost 90% https://www.adexchanger.com/publishers/the-ai-search-reckoning-is-dismantling-open-web-traffic-and-publishers-may-never-recover/ . Many have already shut down. The web is not disappearing. Humans are simply becoming the minority of those who read it. The World That's Coming Imagine the operating system from the movie Her : an ambient, deeply personal assistant that knows you, anticipates you, and quietly handles your relationship with the rest of the world. That is roughly where we are headed, except it is not one assistant; It is dozens, embedded in every device and interface. ChatGPT, NotebookLM, Copilots, and all the next-generation descendants of Google Now. We’re talking in-car assistants, wearables, and ambient surfaces that don’t exist yet. In that world, the web is fully interpreted and curated based on each individual user’s preferences, history, and intent. Visiting a website is rare. Instead the agent reads, ranks, summarizes, personalizes, and presents. The user gets the answer, the recommendation, the purchase, the booking. Your brand appears, if it appears at all, as a fragment inside someone else’s experience. The medium is no longer your website but the agent. Machine Media, Defined Welcome to Machine Media: the era in which agents, not browsers, not feeds, not inboxes, are the interface between your brand and your customer. Your “website” becomes a data source. It could be your site, an API, an MCP server, a structured feed, or any other surface that gets your content into a machine. The “agent” is anything that ingests, curates, and remixes that content for a user. The future of the web is not more traffic to you. It is more curation and reformatting of you, happening everywhere, on surfaces you don’t own and can’t control. The Great Collapse Every earned-media channel that marketers have optimized for two decades, search, social, email, is being absorbed into a single agent-mediated layer. So is commerce, support, booking, contracting, and renewal. Agentic protocols are not just discovery protocols, they are transaction protocols. The agent will not only decide which brand the customer encounters. It will close the sale on the customer’s behalf. The marketing org chart, the agency stack, the martech budget, the attribution model: none of it was built for a world in which one entity interprets, curates, transacts, and decides on the customer’s behalf. The channels still work, but they are rapidly becoming obsolete. A newer technology serves the user better, and that is the only definition of obsolete that has ever mattered. This collapse rewires the economics of the industry. Traditional ad models depend on impressions, clicks, and inventory scarcity, and all three lose meaning when an agent decides what the user sees. Agentic transactions outperform them on every dimension that matters: intent is captured at the moment of decision, the conversion closes inside the same context as the recommendation, and attribution is unambiguous. Early data shows AI-referred traffic converts at roughly twice the rate of traditional search https://ppc.land/over-1-000-brands-now-live-on-chatgpt-ads-via-criteo-as-ai-conversions-near-2x/ in several retail categories. Pay-per-transaction is already emerging as the default monetization model. OpenAI’s Agentic Commerce Protocol https://openai.com/index/buy-it-in-chatgpt/ charges a take rate on completed transactions. Stripe’s Shared Payment Token https://stripe.com/newsroom/news/stripe-openai-instant-checkout infrastructure and Google’s Agent Payments Protocol https://developers.googleblog.com/developers-guide-to-ai-agent-protocols/ are built on the same model. Perplexity has pulled back from advertising https://almcorp.com/blog/perplexity-ai-abandons-advertising-2026-analysis/ entirely. The advertising business is not being replaced by another advertising business. It is being replaced by commerce that closes itself. The New Discipline: Relevance Engineering While SEOs are the best positioned to support this shift, Relevance Engineering is not the next version of SEO. It is the layer that needs to live atop every earned-media discipline going forward: search, content marketing, PR, social, digital partnerships, and even traditional editorial. When agents interpret everything, every earned-media surface is feeding the same intermediary, and every discipline is now in service of a single question: will the agent select, cite, and surface us across the agentic interfaces that matter? Relevance Engineering is the practice of engineering your brand, content, and data to be chosen, cited, and transacted on by machines, across every surface, at every moment of intent. It is the operating layer above the disciplines, not a replacement for any one of them. It is the new operating discipline for the C-suite, and it is what your next CMO should be hired against. Surviving the Machine Media Transition The shift is rapidly emerging, but there are six strategic moves you should be making to prepare. Treat owned content as a data layer, not a destination. Centralize, structure, and make every claim API-first, MCP-ready, and machine-addressable. The website is no longer the product. The product is the data the website happens to render, and any agent should be able to consume it in the format it prefers. Google’s Developer’s Guide to AI Agent Protocols https://developers.googleblog.com/developers-guide-to-ai-agent-protocols/ is the cleanest map of the standards your data needs to be ready for. Adopt the new protocols early. MCP/WebMCP, A2A, UCP, AP2, A2UI, AG-UI, llms.txt, agent-readable schemas, structured feeds, and whatever ships next quarter. These are not optional. They are the rails on which agents discover you, communicate with each other on your behalf, transact, and render the result. Early movers compound, because the models being trained right now are the foundation that will answer queries for the next decade. Late movers do not catch up. They disappear from the answer set. Feed the machines everything, except your moat. Every fact, spec, price, review, policy, and proof point should be available to agents in formats they can ingest, reason over, and cite. The aperture of machine-consumable data you expose is your distribution. The exception is your competitive moat: original research, proprietary data, frameworks you’ve spent years and millions producing. That content gets cloaked or selectively withheld from the LLMs that don’t render JavaScript https://ipullrank.com/cloaking-for-llms , because the same bots that crawl it today are the ones your competitors will use to republish it on their own domain by Tuesday. Measure presence and transaction. Build dashboards for brand visibility inside models and agents, and instrument for the transactions those agents close on your behalf. Traffic is a lagging indicator of a dying medium. Bot activity, rankings for synthetic queries, citation rate, share of model, agentic conversion rate, and revenue closed through agentic protocols are the new scoreboard. Engineer your brand to be agent-legible. Some agents are multimodal and can see your images, your video, and your storefront experience. Most cannot. Even the ones that can render still also leverage structured data, citations, third-party validation, consistent factual presence across the web, and explicit machine-readable claims to decide what is true about your brand. Make yourself legible to every model in every place they look and all the formats they consume. Reorganize around Relevance Engineering. This is not an SEO team’s job. It is a C-suite-owned discipline that spans product, content, data, and brand. Most organizations will need to fold or rebuild three or four existing functions to staff it properly. Do it now, while there is still room to lead. The Stakes Are Higher Than You Think The open web is not coming back. The agents that intermediate everything are being trained now. The cohort of brands that appear in their answers, and that close transactions through them, is being decided now. There is no second wave for laggards in Machine Media. The brands that move first will be the ones the next generation of consumers experiences. The brands that wait will not be punished with lower rankings. They will be unseen and unsellable. The agent will not refer them or transact with them. The agent will simply choose someone else. Either feed the machine or be eaten by it.