News

The future of websites in the agentic AI era: from human showcase to machine data hub

February 10, 2026

The website is splitting in two: a structured data hub for AI agents on one side, a proprietary experience platform for humans on the other. A strategic guide to a structural transformation.

The future of websites in the agentic AI era: from human showcase to machine data hub

Summary

The website as we have known it for 25 years — a digital showcase designed to attract humans, make them scroll, and convert them — is being deconstructed. Not to disappear, but to split in two: on one side, a structured data hub for AI agents (invisible, API-first, machine-readable); on the other, a proprietary experience platform for humans who still visit intentionally. This is not a prediction. It is a process already underway, accelerated by the emergence of UCP (Universal Commerce Protocol), ACP (Agentic Commerce Protocol), WebMCP, and the collapse of the “click” model that has structured the commercial web since its inception.

The question every decision-maker must ask is no longer “How do I optimize my website?” but “If an AI could answer every question my site addresses, who would still come to visit it?”

The click cliff: the classic web is collapsing in numbers

Zero-click is no longer a theory

The data is unambiguous. According to Semrush and SparkToro, 58.5% of Google searches in the United States and 59.7% in the EU now end without any click to an external site. In 2019, this rate was 50%. In 2022, 26% by some measures. The acceleration is exponential.

When Google’s AI Overviews appear in results, the organic click-through rate drops by 61%, falling from 1.76% to 0.61%. Bain & Company reported in February 2026 that 60% of searches are now “zero-click”. E-commerce site traffic has already dropped by 15 to 30% despite stable rankings.

Gartner confirms the structural trend

Gartner predicted as early as 2024 that traffic volume from traditional search engines would drop 25% by 2026, with GenAI solutions becoming “substitute answer engines.” VP Analyst Alan Antin clarified: “The number of searches is not declining. Google is processing more queries than ever. What is changing is what happens after the query.” Gartner even forecasts a 50% drop in organic traffic by 2028.

The illusion of visibility without traffic

A documented case perfectly illustrates this paradox: a site saw its impressions more than double between May 2024 and September 2025, while its click-through rate went from 1.5% to less than 0.5%. More people were seeing its content, but 67% fewer were actually visiting the site. The brand was appearing in AI responses, generating visibility without traffic — a radically new phenomenon.

The paradigm shift: from “Human → Site → Action” to “Human → Agent → Result”

AI agents become the primary users of the web

The web is transitioning from a “pull” model (the human actively searches for information) to a “push” model (the human expresses an intent, the agent executes). In this logic, the website is no longer the final destination — it is one data source among many that the agent consults, synthesizes, and acts upon without ever sending the user there.

As Forbes summarizes: “For 20 years, we have treated our websites as the undisputed centers of our digital universe. That universe is about to collapse.” The concept of “zero-click commerce” is its concrete manifestation: a consumer asks their AI agent to find running shoes under €150, eco-friendly, size 42. The agent evaluates reviews, compares specs, checks stock via API, and presents three options — all without ever visiting a website.

Agentic commerce: the protocols accelerating disintermediation

Google’s announcement of the Universal Commerce Protocol (UCP) at NRF 2026, co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, formalizes this transformation. UCP enables any AI agent to:

  • Discover products via structured feeds (ucp_discovery)
  • Negotiate merchant capabilities (ucp_negotiate)
  • Link user identity securely (ucp_identity)
  • Complete the purchase without leaving the AI interface (ucp_checkout)

On the OpenAI/Stripe side, the Agentic Commerce Protocol (ACP) does the same for conversational surfaces like ChatGPT. Most large retailers will need to support both protocols.

McKinsey projects $3 to $5 trillion in global agentic commerce by 2030. This is not a niche market — it is the next web.

The website is not dying — it is splitting in two

Role #1: the fundamental data hub (for machines)

The first incarnation of tomorrow’s website is invisible to humans. It is a silent, structured, machine-readable layer whose “product” is verifiable data delivered via API.

The concrete implications:

  • API-first: AI agents interact with your API, not your graphical interface. A company without a robust, well-documented public API risks becoming invisible in the agentic ecosystem.
  • Structured data: Schema.org, JSON-LD, FAQPage, HowTo, Product markup — these are no longer “SEO bonuses,” they are the condition for existence in AI responses.
  • Merchant Center feed: for commerce, the Google Merchant Center product feed becomes the primary source through which AI interprets a merchant’s offering.
  • Protocol interoperability: UCP, MCP (Model Context Protocol), A2A (Agent-to-Agent), AP2 (Agent Payments Protocol), OAuth 2.0 — the site must “speak” these languages.

Aaron Levie, CEO of Box, anticipates 1,000x growth in API usage in the agentic era. The parallel is striking: just as “mobile-first” was a matter of survival, “API-first” is the new baseline.

Role #2: the proprietary experience platform (for humans)

The second role of the website is equally crucial — but radically different. If AI can answer all informational questions, then the website must become a place people come to voluntarily, not because Google sent them there.

What this implies:

  • Immersive experiences: mini-games, 3D, interactive storytelling, gamification — interactive elements increase engagement by 100 to 150% and social sharing by 22%.
  • AI hyper-personalization: the site dynamically reconfigures according to visitor intent, using AI to reorder content, adapt layouts in real time, and deliver on-site conversational experiences.
  • Proprietary agents: the main threat is being reduced to a “faceless fulfillment node” in a marketplace controlled by third-party agents. The solution: deploy proprietary AI agents, trained on your first-party data, capable of offering personalized advice that a generic shopping agent cannot replicate.
  • The antidote to commoditization: if your site merely sells products available everywhere, the agent will choose the cheapest. If your site offers a unique experience — vertical expertise, community, exceptional post-purchase service — it becomes irreplaceable.

The new interfaces of the agentic web

Cloudflare “Markdown for Agents”: making the web readable

Announced on February 12, 2026, Cloudflare proposes to automatically serve an optimized Markdown version of any website to AI crawlers, directly from the CDN. The ambition is to become the default reading gateway for agentic traffic — and to define the metrics the market will adopt: “agent impressions,” “markdown served,” “token footprint.”

WebMCP: making the web actionable

Incubated at the W3C and available in preview in Chrome since February 2026, WebMCP allows websites to expose structured tools (with name, description, input schema, execution callback) that AI agents can invoke directly in the browser.

The fundamental difference from classic scraping: WebMCP transforms sites into declarative APIs in the browser, with the browser as arbiter of permissions, identity, and user visibility. Agents no longer need to “pretend to be humans clicking buttons” — they call typed, secure, auditable tools.

The strategic implication is profound: sites that expose the cleanest and most stable tool contracts will be those agents prefer, as they minimize failures and ambiguity. This is not a DX (developer experience) issue — it is a distribution issue.

Comparative table: the two new interfaces

The web economy is reconfiguring

The advertising model crisis

DimensionCloudflare Markdown for AgentsWebMCP
FunctionMake the web readable by agentsMake the web actionable by agents
LayerCDN / edgeBrowser (browser API)
MechanismAutomatic conversion to MarkdownExposure of structured tools (registerTool)
SecurityServer-side controlHuman-in-the-loop, browser permissions
Use caseReading, extraction, synthesisPurchase, booking, form submission
MaturityBeta (paid Cloudflare plans)Early preview Chrome / W3C

AI agents do not see banners, do not watch pre-rolls, do not click on sponsored links. The advertising model based on human attention is structurally incompatible with an agentic ecosystem. For publishers and any business dependent on advertising revenue, this is an existential threat.

New monetization models

The machine-to-machine (M2M) economy is emerging with new levers:

  • Premium API access: pricing by volume/quality of data consumed by agents
  • Data licensing: monetizing proprietary data as “fuel” for LLMs
  • Verification-as-a-Service: certification of data authenticity and freshness
  • Agent marketplace: specialized vertical agents become marketable products
  • Native advertising in AI responses: Google is already integrating promotions in AI Mode

New KPIs

Traditional metrics (sessions, page views, bounce rate) are becoming insufficient. The new indicators include:

What this means for different types of sites

E-commerce sites

Old KPINew KPI
Sessions / organic trafficAI Visibility Score (frequency of appearance in AI responses)
Click-through rate (CTR)Multi-platform citation rate
Google rankingsBrand Mention Sentiment (context and tone)
On-site conversion rateAgent Task Success Rate (agentic transaction success rate)
Page viewsQuestion Coverage Rate (coverage of your audience’s questions)

The most direct and urgent impact. The classic e-commerce site (catalog → cart → checkout) is threatened with total disintermediation by shopping agents. The strategy: implement UCP/ACP, optimize product feeds, expose data via API, and build an irreplaceable proprietary experience (expertise, community, service).

Shopify already natively integrates UCP — Shopify merchants have automatic access to Google’s AI surfaces. For other platforms, integration requires significant technical work.

Content sites / publishers

The most vulnerable to zero-click. Pure informational sites suffer the most severe traffic drops (up to -69% click-through rate for informational queries with AI Overviews). The strategy: pivot toward original research (proprietary data that AI cannot find elsewhere), interactive experiences, and Answer Engine Optimization (AEO) to maximize citations.

SaaS / B2B sites

Informational traffic will decline, but the remaining visitors will be more qualified — pre-filtered by AI. The strategy: become the source AI cites and recommends (thought leadership, original data, E-E-A-T), while exposing your capabilities via API/MCP so agents can interact with your product directly.

Showcase sites / SMEs

Static “brochure” sites are the first victims. If your site merely displays information AI can summarize, no one will come. The solution: structured, clear, explicit content about who you help, how, and why you can be trusted — optimized to be understood by humans, search engines, and AI systems alike.

7-step action plan

1. “Agent-First” audit of your digital assets

Evaluate all your digital assets through the lens of machine readability. Map your critical data sources, evaluate their accuracy and freshness, and identify the path to exposing them via API.

2. Adopt an API-First philosophy

Treat your APIs as primary products, not internal technical tools. Invest in DX (developer experience) with clear documentation — the “users” of your APIs will be other developers’ agents.

3. Implement structured data at scale

Schema.org, JSON-LD, FAQPage, HowTo, Product, LocalBusiness — structured markup is the condition for existence in AI responses. Every page must be machine-readable.

4. Optimize for AEO (Answer Engine Optimization)

Beyond classic SEO, optimize to be cited by AIs. Clear, definitive answers at the start of sections, logical structure, attributable sources, authority signals (E-E-A-T).

5. Prepare UCP/ACP integration (if commerce)

For commerce: join the UCP waitlist via Google Merchant Center, ensure your product feeds are complete and up to date, and plan ACP integration if you use Stripe.

6. Build your proprietary experience platform

Invest in what AI cannot replicate: interactive experiences, proprietary vertical agents, community, unique expertise. The site must become an intentional destination, not a mere transit point.

7. Redefine your KPIs and train your teams

Move from traffic-based reporting to AI visibility-based reporting. Train your teams in new skills: prompt engineering, AI governance, data engineering, AIO (Agent Interface Optimization).

Horizon 2027-2030: toward the post-interface web

Agents negotiate with agents

The ultimate endpoint is not simply humans using agents — it is agents interacting with other agents. The customer’s AI agent negotiates directly with the merchant’s AI agent: price, availability, delivery, personalization, after-sales service — all autonomously. This is the emergence of the machine-to-machine economy, where transactions become computational workflows rather than human-driven exchanges.

The Gartner prediction for 2027

By 2027, autonomous agents will operate as “digital employees” with assigned objectives, executing complex multi-step tasks with minimal supervision. Adoption in retail and e-commerce will reach 70-85% penetration.

The web is not dying — it is mutating

The website is not disappearing. It ceases to be the center of gravity of digital strategy to become one node among many in a distributed ecosystem. The most accurate historical parallel is not the “death of the website” — it is the transformation that physical stores underwent with the arrival of e-commerce. Stores still exist. But their role, design, and economics have been fundamentally redefined.

The same mutation is occurring for the web. The sites that will survive and thrive will be those that have understood they must now serve two radically different audiences: machines that want structured data and reliable APIs, and humans who want experiences that only intentional visits can provide.

References

    The future of websites in the agentic AI era - Busony