TL;DR: Google has officially told developers to treat AI agents as a distinct website visitor alongside humans. Their new guidance lays out seven rules for “agent-friendly” websites that overlap almost entirely with Answer Engine Optimization (AEO) principles. For businesses investing in content, this means the same structural work that gets you cited by ChatGPT and Perplexity also makes your site navigable by AI agents that browse, compare, and buy on behalf of users.
LA & CO Content Agency breaks down what changed, what the seven rules are, and exactly what your business should do about it.
What Did Google Actually Announce About Agent-Friendly Websites?
Google published an official guide on its developer hub (web.dev) titled “Build agent-friendly websites” that advises developers to design for AI agents as a new visitor class alongside human users.
The guide, authored by Kasper Kulikowski and Omkar More, was last updated on April 1, 2026.
It opens with a statement that frames the shift clearly: websites now have “a new type of visitor” because users are moving from manual browsing to delegating tasks to AI agents.
This is not a blog post or an opinion piece.
It sits on web.dev alongside Google’s official guidance on performance, accessibility, and web standards.
That placement matters because it signals that Google considers agent-friendliness a core web development discipline, not a trend.
The timing is also significant.
Google published this guidance weeks before Google I/O 2026 (May 19-20), where Chrome is scheduled to present updates on browser-based agent interactions.
Alongside the guide, Google launched the WebMCP early preview program, a proposed web standard for structured agent-website interaction.
The data supports the urgency.
A 2026 report from HUMAN Security found that AI agent traffic grew 7,851% year-over-year.
According to the Imperva Bad Bot Report, automated traffic (including AI agents) accounted for 51% of all internet traffic in 2025.
And 2.3% of all AI agent activity now occurs on checkout pages, a clear signal that agents are not just reading websites but actively making purchases on behalf of users.
For businesses that rely on organic search and content marketing, this changes the equation.
Your website now needs to serve two audiences simultaneously: the human who reads your content and the AI agent that navigates, compares, and transacts on their behalf.
How Do AI Agents Actually See Your Website?
AI agents interpret websites through three channels: screenshots, raw HTML, and the accessibility tree.
Modern agents combine all three to navigate, and the quality of what they find determines how well they can act on a user’s behalf.
Screenshots
The agent takes a snapshot of the rendered page and uses a vision model to identify elements.
It can recognize that a search bar at the top right is likely a global search, or that a large red button labeled “Delete” should be treated with more caution than a small “Help” link.
However, screenshot analysis is slow and expensive in processing terms.
Agents use it as a backup when the page structure is confusing, not as their primary method.
Raw HTML
The agent reads the DOM directly.
It understands how elements are nested, the logical hierarchy of the page, attributes like IDs and classes, and the raw text that forms the page’s content.
This is how the agent determines relationships between elements.
If a “Buy Now” button sits inside a product container, the agent knows that button belongs to that specific product.
The Accessibility Tree
This is where it gets interesting for content strategists.
The accessibility tree is a browser-native API that distills the DOM into its most essential components: the roles, names, and states of interactive elements.
Google’s guide calls it a “high-fidelity map” that strips away the visual noise of CSS to focus on pure functionality.
For AI agents, the accessibility tree is the cleanest signal available.
It tells the agent exactly what every toggle, slider, input field, and button does without needing to interpret visual design.
Why Combined Modalities Matter
No single channel gives the full picture.
In the DOM, an agent might see a <div> without knowing it has been styled to function as a button.
In a screenshot, the agent might see where that button sits visually but not know what it does when clicked.
Modern agents cross-reference all three channels to fill these gaps.
The takeaway for businesses: your job is to provide clean, consistent signals across all three.
If your content is buried in JavaScript widgets, hidden behind complex hover states, or trapped in image-only formats, agents cannot read it.
This is the same problem AI crawlers from ChatGPT, Perplexity, and Google’s AI Overviews face when deciding whether to cite your content.
What Are Google’s Seven Rules for Agent-Friendly Websites?
Google’s checklist covers seven specific recommendations, and nearly all of them overlap with existing web accessibility best practices.
If your site already follows WCAG guidelines, you are most of the way there.
| Rule | What It Means | You’re Already Covered If… |
|---|---|---|
| Use semantic HTML | Use <button> and <a> tags instead of styled <div> and <span> elements. Agents recognize semantic elements as interactive. |
Your site uses proper HTML elements rather than CSS-styled containers for buttons and links. |
| Keep layouts stable | Avoid shifting element positions between pages. An “Add to Cart” button that moves to a different location on each product page confuses screenshot-based agents. | Your templates are consistent and buttons appear in the same positions across similar pages. |
Set cursor: pointer |
Apply pointer cursor styling to clickable elements. Google calls this “a strong signal for actionability.” | Your CSS framework already applies pointer cursors to interactive elements (note: Tailwind v4 removed this default for native buttons). |
| Link labels to inputs | Add the for attribute on <label> tags to connect them to form inputs. This tells agents exactly what each field is for. |
Your forms follow basic accessibility standards with properly linked labels. |
| Minimum interactive size | Ensure clickable elements have a visible area larger than 8 square pixels. Smaller elements get filtered out by visual analysis. | You design for mobile touch targets (which are typically much larger than 8px). |
| No ghost overlays | Avoid transparent overlays that cover interactive elements. Agents may discard elements that appear covered, even if the overlay is invisible to humans. | Your site does not use overlay-heavy popup patterns or invisible blocking layers. |
| Visible state changes | All actions should be clearly reflected in the interface. When a user (or agent) clicks a button, the result should be visually confirmed. | Your buttons show loading states, success confirmations, or other visual feedback on interaction. |
The most striking observation about this list is that it contains almost nothing new for accessibility practitioners.
Every rule maps to an existing WCAG recommendation.
Semantic HTML is the foundation of WAI-ARIA Authoring Practices.
Label linking, minimum target sizes, and avoiding hidden interactive elements are all standard accessibility requirements.
What is new is the framing.
Google is no longer just telling developers to build accessible websites for people who use screen readers.
They are telling developers to build accessible websites for AI agents that browse, compare, and purchase on behalf of millions of users.
The audit is the same.
The business case just doubled.
Why Does This Matter for Content Strategy, Not Just Developers?
Google’s guidance validates that structured, semantic, machine-readable content is now a business requirement, not a developer nice-to-have.
The same principles that make your site navigable by AI agents also make your content citable by AI search platforms.
Consider the overlap.
AEO (Answer Engine Optimization) requires clean HTML structure, semantic markup, self-contained sections, and machine-readable data so that platforms like ChatGPT, Perplexity, Claude, and Gemini can extract and cite your content.
Agent-friendly design requires clean HTML structure, semantic markup, stable layouts, and machine-readable interactive elements so that AI agents can navigate and act on your site.
The structural foundation is identical.
Google’s own guide states it directly: “Everything we suggest to make a site ‘agent-ready’ also makes sites better for humans.”
This is also why the dual-layer principle matters.
Interactive elements on your website (calculators, comparison tools, product finders) need static HTML fallbacks because AI agents and AI crawlers cannot process JavaScript.
The human visitor gets the rich, interactive experience.
The AI agent and the AI search crawler both get the same underlying data in a format they can read.
There is an important distinction in how structured data fits into this picture.
Schema.org markup (Article, FAQ, Product, HowTo) covers the “nouns” of your website: this is a product, this is a price, this is a review, this is a frequently asked question.
WebMCP, Google’s proposed new standard, will cover the “verbs”: buy this product, search this category, book this appointment.
A complete AI visibility strategy needs to address both layers.
Content strategy is no longer just about what you write.
It is about how that content is structured, marked up, and made accessible to every type of visitor, whether human, AI crawler, or AI agent.
What Is WebMCP and Should You Care About It Yet?
WebMCP (Web Model Context Protocol) is a proposed web standard from Google and Microsoft that lets websites expose structured, callable actions directly to AI agents.
It is currently in early preview through Chrome’s developer program, and signups are open.
The simplest way to understand WebMCP is through a comparison.
Schema.org tells AI systems what things are: this is a product, this is a price, this is an event.
WebMCP tells AI agents what things do: buy this product, search this category, submit this form.
Without WebMCP, an AI agent trying to complete a purchase on your site has to load your CSS, parse the accessibility tree, identify the “Add to Cart” button, and simulate a click.
With WebMCP, the agent can call a structured function directly, like an API call, bypassing the visual interface entirely.
There are two implementation modes.
Declarative mode uses HTML attributes for simple, predictable interactions like search forms and newsletter signups.
Imperative mode uses a JavaScript API for complex, stateful interactions like ecommerce checkout flows where available actions change based on what the user has done.
For most businesses right now, WebMCP is the horizon, not the immediate action item.
The standard is still experimental.
It may change significantly before reaching stable release.
But the direction it signals is clear: Google is building toward a web where AI agents can interact with websites as smoothly as humans do.
The businesses that will be ready for that future are the ones investing in the structural fundamentals today: semantic HTML, clean content architecture, schema markup, and AEO-optimized content.
Watch Google I/O on May 19-20 for potential updates on WebMCP’s roadmap and Chrome’s agent interaction features.
What Should Your Business Do Right Now?
Start with a combined audit that covers both AI citation readiness and agent-friendliness, because the checklist is essentially the same.
Content Layer (AEO)
Structure every page with an answer-first approach, placing a direct response to the reader’s question in the opening 40 to 60 words.
Use natural-language question headings that match what real people type into ChatGPT, Perplexity, and Google’s AI Overviews.
Build self-contained sections that AI platforms can extract and cite independently without needing context from the rest of the page.
Include comparison tables and structured data wherever the topic supports it, because tables are among the most extractable content formats for AI.
Implement FAQ, Article, Product, and HowTo schema markup on every relevant page.
Technical Layer (Agent-Friendly)
Pull your top five highest-traffic pages and run them against Google’s seven rules.
Run a WCAG-AA accessibility scan using Lighthouse, axe DevTools, or the WAVE browser extension.
The overlap with agent-friendliness is near-total, so one scan covers both.
Replace any styled <div> elements functioning as buttons with proper <button> and <a> tags.
Ensure every interactive element (calculators, tools, dynamic filters) has a static HTML fallback that AI crawlers can read.
Check that all form labels are properly linked to their inputs using the for attribute.
Strategic Layer
Test your brand’s visibility across ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews by asking the questions your customers ask.
Track AI-referred sessions in your analytics platform, because this traffic segment grew 527% in just six months during H1 2025.
Watch Google I/O (May 19-20) for WebMCP updates, Chrome agent features, and any signals about how agent-friendliness may factor into search visibility.
If you do not have the in-house expertise to run this audit, request a free AI visibility audit from LA & CO and we will assess your site’s readiness across all three layers.
How Does Agent-Friendly Design Connect to Answer Engine Optimization?
Agent-friendly design and AEO are two expressions of the same underlying principle: making your website’s content and functionality machine-readable.
Agent-friendly design optimizes the navigation layer, ensuring AI agents can browse, interact with, and transact on your site.
AEO optimizes the citation layer, ensuring AI search platforms can extract, reference, and recommend your content.
Both require the same structural foundation: semantic HTML, structured data, clean content hierarchy, self-contained sections, and machine-readable markup.
This is why accessibility practitioners who have been doing the work for years are already most of the way to passing Google’s new agent-friendly checklist.
The structural discipline is identical.
The audience has expanded.
Think of it as an evolution, not a replacement.
Good SEO builds the foundation of keyword targeting and technical health.
Great AEO adds the answer-first structure, entity signals, and schema markup that get your content cited by AI platforms.
Agent-ready design adds the interactive layer that lets AI agents navigate and transact on your site.
Each layer builds on the one before it.
The businesses that will lead in AI visibility are those investing across all three layers simultaneously, because the same piece of well-structured content serves every audience: the human reader, the AI search crawler, and the AI agent acting on a user’s behalf.
LA & CO Content Agency builds every piece of content on this multi-layer framework. Get a free quote to see how AEO-optimized, agent-ready content can work for your business.
