Agentic navigation in PageSpeed Insights: what it means for websites prepared for AI

Google PageSpeed Insights has started to show a new analysis area called Agentic navigation, a category still under development that aims to assess whether a website is ready to be understood, explored and used by artificial intelligence agents. This update appears at a time when the digital experience is no longer designed only for people, but also for automated systems capable of searching, interpreting content, filling in forms, comparing products, checking information and interacting with interfaces on behalf of users.

Until recently, when discussing the technical performance of a website, attention was mainly focused on metrics such as loading speed, visual stability, accessibility, best practices and SEO. These pillars remain essential. However, the arrival of Agentic navigation in PageSpeed Insights shows that the quality of a website is also beginning to be assessed by its ability to communicate clearly with AI agents. The question is no longer only: is the website fast and easy to use? It is also: is the website structured well enough for a machine to understand what exists on each page and act correctly?

What does Agentic navigation mean?

The expression Agentic navigation refers to a website’s ability to allow digital agents, powered by artificial intelligence, to move through pages autonomously, understand content structure and perform tasks reliably. These agents may be personal assistants, systems integrated into browsers, advanced search tools, commercial comparison bots or applications that help users make decisions.

In practice, an AI agent needs to understand whether it is looking at a menu, a form, a product list, a purchase button, a privacy policy, a pricing table or a contact page. The clearer the technical and semantic structure of the website, the greater the probability that the agent will correctly interpret the purpose of each element.

This category is still evolving, but its appearance in PageSpeed Insights is a relevant signal. Google is indicating that the web of the future will need to be more readable for both humans and intelligent systems. This brings together areas such as accessibility, structured data, information architecture, performance, security and user experience.

Why is this update important?

For years, companies invested in websites optimised for people and search engines. The goal was to ensure that users could find information, navigate without friction and convert. At the same time, search engines needed to crawl pages, index content and understand the relevance of each URL.

Now, a third layer is entering the picture: AI agents. These systems do not simply index a page. They may try to interact with it. They may read a product description, compare options, identify a button, fill in fields, submit requests or collect data to support a decision. If the website is poorly structured, the agent may fail, misinterpret the page or abandon the task.

For brands, online stores, professional services and digital platforms, this represents an important shift. A website that is not understandable to AI agents may lose opportunities for visibility, recommendation and conversion in increasingly automated digital ecosystems.

The passed audits: accessibility tree and CLS

In the PageSpeed Insights screenshots, the category shows a score of 2/2 in passed audits. Among the validated elements are a well-formed accessibility tree and a Cumulative Layout Shift score of zero.

The accessibility tree is a technical representation of the page that helps assistive technologies and automated systems understand the elements available. When this tree is well formed, menus, buttons, headings, forms and content become clearer for screen readers and AI agents. This confirms an idea we have defended for a long time: accessibility is not only an ethical or legal responsibility; it is also a technical advantage.

Cumulative Layout Shift, known as CLS, measures the visual stability of a page. A score of zero indicates that visible elements do not unexpectedly change position during loading. For a human user, this avoids accidental clicks and frustration. For an AI agent, it reduces the risk of interacting with the wrong elements. If a button moves while the page is loading, an automated action may become less reliable.

The role of WebMCP

The new Agentic navigation area also mentions WebMCP, a concept associated with the creation of structures that help AI agents understand and interact with web pages. In the non-applicable audits presented, there are references to WebMCP form coverage, registered WebMCP tools, valid WebMCP schemas and the llms.txt file.

Although some of these checks still appear without a score, they clearly point in one direction: websites will need to declare their intentions, functionalities and permissions more clearly. Just as structured data helps search engines understand products, recipes, events or articles, new mechanisms designed for AI agents may help intelligent systems understand which actions they can perform and in what context.

For companies, this means that the technical preparation of a website should no longer be limited to what is visible on the screen. Code, semantics, attributes, guidance files and information organisation are becoming even more strategic.

The llms.txt file and the new guidance layer for AI

One of the most interesting references in this new category is the llms.txt file. Just as robots.txt guides search engine crawlers, llms.txt is designed as a way to help large language models understand how they should interpret or use the information on a website.

This file may become relevant for brands that want to indicate priority content, contextualise their activity, organise documentation or explain how their data should be read by AI systems. There is still room for evolution, but the presence of this recommendation in Google’s analysis environment shows that the relationship between websites and language models will become increasingly structured.

For a company, this raises important questions: which content should be easily understood by AI? Which pages best represent the brand? Which information should be protected? Which data should be made available for automated consultation? These decisions combine digital marketing, technology, content and strategy.

Impact on SEO and organic visibility

Agentic navigation does not replace traditional SEO, but it may become a natural extension of technical optimisation. A website well prepared for AI agents tends to have a good heading structure, semantic HTML, clear navigation, organised content, accessible forms and low visual instability. All of these factors already contribute to a better experience and can support organic performance.

Modern SEO is no longer limited to keyword repetition. It involves search intent, authority, technical quality, structured data, speed, accessibility and the ability to respond to users’ real needs. With AI, a new dimension is added: the ability of a page to be interpreted by systems that do not see the interface like a person, but still need to understand its logic.

In the future, AI assistants are likely to play a growing role in the discovery of products, services and content. When a user asks for recommendations, comparisons or actions, agents may favour websites that provide clear, structured and trustworthy information. For this reason, investing in this preparation is also a way to protect a brand’s digital visibility.

The impact on e-commerce

In e-commerce, Agentic navigation can have a direct impact on conversion. Imagine an AI agent tasked with finding a product, comparing features, checking availability, analysing shipping costs and starting a purchase. For this to happen without errors, the online store must present products, variants, prices, buttons, forms and policies clearly.

A store with confusing HTML, poorly accessible filters, buttons without clear labels or ambiguous forms may make it difficult for digital agents to act. By contrast, a well-structured store allows these systems to recognise categories, product attributes, reviews, availability and the steps in the purchasing process.

For platforms such as Shopify, this is particularly relevant. Online stores already benefit from solid ecosystems, but theme customisation, integrations and apps can introduce technical problems if they are not implemented correctly. AI readiness should be part of quality audits, especially in projects that depend on organic traffic, paid campaigns and direct conversion.

Forms, actions and trust

Forms are one of the most sensitive areas in this new logic. Contact forms, quote requests, registrations, bookings, simulations and checkout processes depend on clear fields and understandable validations. If an AI agent has to fill in a form on behalf of a user, it needs to know which field corresponds to name, email, phone number, address, message or consent.

Fields without labels, generic buttons, unclear error messages and inconsistent validations create barriers. The same applies when forms are visually appealing but technically weak. The human experience may seem acceptable, but automated reading may fail.

Trust will also be decisive. AI agents will need to understand whether an action is safe, whether it involves payment, whether it requires consent or whether it sends personal data. This forces companies to make their interfaces clearer and align technology, privacy and communication.

Accessibility as the foundation of an AI-ready web

The relationship between accessibility and AI agents is direct. An accessible page is, by nature, more structured, more semantic and easier to understand. Properly ordered headings, alternative text, identifiable buttons, keyboard navigation and labelled forms help people with different needs and also support automated systems.

For this reason, the new PageSpeed Insights category reinforces something essential: accessibility should not be treated as a final layer of a project. It should be present from the planning stage of architecture, design, development and content production.

When a page is built with good practices, every stakeholder benefits. The user finds what they need with less effort. The search engine interprets the content better. The AI agent navigates with more precision. The brand gains consistency, trust and adaptability.

How to prepare a website for this new reality?

Preparation for Agentic navigation begins with a careful technical audit. It is necessary to assess performance, visual stability, accessibility, semantic HTML, content structure, structured data, forms, navigation, policies and guidance files.

It is also important to review third-party components. Many websites use plugins, apps, chat widgets, consent tools, advertising scripts and external integrations. These elements can affect performance, create visual instability or make automated page reading more difficult.

Another essential point is documentation. A brand should have clear pages about services, products, contacts, terms, privacy, deliveries, returns and frequently asked questions. The more transparent the information is, the easier it will be for intelligent systems to interpret the company’s value proposition.

What should companies do now?

Even if the category is still under development, companies should not wait for full maturity before acting. The history of the web shows that brands that adapt early to new technical requirements gain a competitive advantage. This happened with mobile, HTTPS, Core Web Vitals, structured data and accessibility.

The first step is to look at PageSpeed Insights not only as a speed tool, but as a digital quality panel. A good performance score is important, but it is not enough. It is necessary to understand what each metric reveals about the real user experience.

The second step is to involve marketing, development, content and UX teams. Preparing for AI agents is not only a technical responsibility. It depends on how information is written, organised and presented. It depends on design decisions. It depends on the site architecture. It depends on the technology used.

The third step is to create a continuous improvement process. Websites are not closed projects. They change with campaigns, new pages, new integrations, new tools and new commercial goals. Each change can improve or harm readability for humans, search engines and AI agents.

A new phase of the digital experience

Agentic navigation represents a new phase of the digital experience. For a long time, websites were optimised for screens. Then, they were optimised for mobile devices. Later, for performance and Core Web Vitals. Now, we are starting to optimise for digital entities that can act on behalf of users.

This does not mean that websites are no longer made for people. Quite the opposite. The principles that help AI agents are, to a great extent, the same ones that make the human experience better: clarity, structure, speed, stability, accessibility and trust.

The difference lies in the level of demand. A user may interpret a poorly written button through visual context. An AI agent needs more explicit signals. A user may understand that a field is for email because it is next to another element. An automated system needs labels, structure and validation. A user may work around a confusing page. An agent may simply fail.

A new opportunity

The new Google PageSpeed Insights data on Agentic navigation shows that the web is moving towards a more intelligent, automated and demanding reality. Fast and beautiful websites will continue to matter, but their quality will increasingly be measured by their ability to be understood by people, search engines and AI agents.

For companies that depend on digital channels, this is an opportunity to review technical foundations, improve accessibility, strengthen SEO, prepare content and make platforms more reliable. Artificial intelligence does not reduce the importance of good websites. On the contrary, it increases the need for well-built sites.

BYDAS helps brands prepare websites, online stores and digital strategies for this new phase, combining web development, performance, technical SEO and user experience focused on results.