Is your website ready for AI agents?

For years, websites have largely been designed around two audiences.

People.

And search engines.

That is starting to change.

As AI assistants become capable of doing more than answering questions, websites may increasingly need to support a third type of visitor: AI agents that can navigate pages, interact with forms and complete tasks on behalf of users.

Google Chrome is already experimenting with this through a new Agentic Browsing category within Lighthouse.

The new checks look at how easily AI agents can understand and interact with a website, combining familiar website fundamentals such as accessibility and layout stability with emerging technologies including WebMCP and llms.txt.

This does not mean every organisation suddenly needs to rebuild its website for AI agents.

The technology is still experimental.

But it does provide another useful signal about where digital experiences may be heading.

Good websites increasingly need to be understandable not only by people and search engines, but potentially by the systems acting on their behalf.

What is an AI agent?

Most people have become familiar with generative AI through tools that answer questions.

You ask something.

The AI provides a response.

AI agents take that idea further.

Instead of simply providing information, an agent can potentially carry out actions for the user.

Imagine asking an AI assistant:

“Find an appointment next Thursday afternoon and book it for me.”

Or:

“Find a suitable volunteering opportunity near me and help me apply.”

Or:

“Find a hotel in London within my budget and complete the booking.”

For this to work reliably, the AI needs to do more than understand the written content on a website.

It needs to understand how the website itself works.

Which field is asking for a person’s name?

Which button submits the form?

What does a particular dropdown control?

What happens after somebody clicks confirm?

And which actions is the website actually allowing the agent to take?

Chrome refers to the process of an agent interacting with website controls as agent actuation.

Traditionally, an agent may need to interpret a website and simulate the same clicks and text entry a human would make.

Newer approaches are looking at whether websites can communicate those available actions more explicitly.

What is Lighthouse Agentic Browsing?

Google Lighthouse is already widely used by developers, marketers and website owners to assess areas including performance, accessibility, SEO and technical best practice.

Chrome’s experimental Agentic Browsing category adds a new group of checks focused specifically on machine interaction.

At the moment, Google says the category examines areas including:

Accessible names, labels and the integrity of the website’s accessibility tree.

Cumulative Layout Shift, which measures whether page elements move unexpectedly.

WebMCP tools and whether their structure has been implemented correctly.

The availability of an llms.txt file as an emerging machine-readable discoverability signal.

Unlike Lighthouse Performance, however, there isn’t currently an Agentic Browsing score out of 100.

Google describes the category as experimental and unbenchmarked.

It currently reports a fractional pass rate alongside individual warnings and pass or fail results.

That distinction matters.

This is not a new Google ranking factor that organisations need to urgently optimise towards.

It is a testing framework designed to provide actionable signals while the standards around agentic browsing continue to develop.

Why accessibility matters for AI agents

One of the most interesting aspects of Google’s Agentic Browsing work is the importance placed on accessibility.

AI agents can use a website’s accessibility tree as a machine-readable representation of the page.

That means some of the same principles that make websites easier for people using assistive technologies can also make interfaces easier for machines to understand.

An interactive element should have a clear programmatic name.

Form fields should have appropriate labels.

Buttons should clearly communicate their purpose.

HTML should be structured semantically.

Relationships between different elements should make sense.

None of this is new.

These are already principles of good accessible web development.

And accessibility should always be approached as a requirement for people first.

But it is interesting that the same foundations are also becoming important as websites begin interacting with a new generation of AI systems.

We discussed the wider importance of accessibility, clear journeys and website structure in our guide to what a charity website should include.

The broader lesson applies well beyond charities.

A website that clearly communicates what each element does is easier for everyone to use.

Stable layouts matter when an AI agent is clicking for you

Most people have experienced a website where something suddenly moves while the page is loading.

You go to click one button.

An image, advert or banner appears.

The layout shifts.

You accidentally click something else.

That is frustrating for a human visitor.

It can also make automated interaction less reliable.

Google therefore includes Cumulative Layout Shift, or CLS, within its current Agentic Browsing checks.

If an agent identifies an element in one position and that element moves before the interaction takes place, there is greater potential for the wrong action to be triggered.

Again, this isn’t an AI-specific problem.

Layout stability already matters for user experience and forms part of Google’s Core Web Vitals.

Agentic browsing simply introduces another reason to care about it.

A properly maintained website should behave predictably.

That is one reason ongoing WordPress support and maintenance is about considerably more than installing plugin updates.

Performance, compatibility, layout stability and technical quality all contribute towards how reliably people and systems can use a website.

What is WebMCP?

WebMCP is probably the least familiar part of Google’s experiment.

It is a proposed web standard designed to allow websites to expose structured tools to AI agents.

In plain English, it gives a website a way to explicitly describe some of the actions an agent can carry out.

For example, a website might expose an action such as:

book_appointment

submit_application

search_products

or:

add_to_cart

The website can provide information explaining what that action does, what information is required and how the agent should interact with it.

Chrome’s WebMCP documentation describes the technology as a way of exposing structured tools for AI agents so that website interactions can become more accurate and reliable.

This potentially reduces the amount of guessing an agent needs to do.

Instead of:

Look at page → identify form → interpret fields → find button → attempt action.

The website could provide a structured description of the action that is available.

That is an important change.

What could WebMCP mean in practice?

The concept becomes more interesting when it is applied to real website journeys.

A healthcare organisation might allow an agent to interact with an appointment-booking process.

A charity could eventually make some volunteer application journeys easier for compatible agents to understand.

A retailer could expose product search or filtering tools.

A restaurant might support a structured reservation action.

A professional services organisation could make it easier for an agent to book a consultation.

An ecommerce website might expose actions that help an agent find and purchase an appropriate product.

Google’s own WebMCP guidance uses examples around purchases and other critical user journeys when explaining where structured agent interactions could be useful.

Historically, a large proportion of website optimisation has focused on making an organisation easier to find.

Agentic browsing introduces another question.

Once an AI system finds the website, can it successfully use it?

AI agents and AI search are not the same thing

This distinction is particularly important.

There is currently a great deal of discussion around optimising websites for AI search.

Can Google AI features understand your content?

Can ChatGPT or another AI platform discover your organisation?

Does your website demonstrate genuine expertise?

Can machines clearly understand what your organisation does?

These questions largely relate to discovery, retrieval and understanding.

That is where strong SEO, technical structure, useful content and clear information architecture continue to matter.

Chrome itself makes this distinction.

When agents are searching the web, established SEO principles still apply.

Agentic Browsing is instead looking at what happens when an agent arrives on the website and tries to interact with it.

That gives us two increasingly important questions:

Can an AI system find and understand your organisation?

And:

Can an AI system successfully use your website?

We explore the first of those areas in more detail throughout our work around AI search and website visibility.

If you’re unsure how well your own website currently performs across technical discoverability, content clarity, AI search readiness, accessibility and other important areas, you can check your website using our free Website Visibility Checker below.

What about llms.txt?

The inclusion of llms.txt within Agentic Browsing has understandably attracted attention.

An llms.txt file is an emerging convention intended to provide LLMs and AI agents with a machine-readable summary of important website content.

Chrome’s Agentic Browsing audit checks whether one is available at the root of the website.

However, this needs some context.

Google currently describes llms.txt as optional within this Lighthouse audit.

If an llms.txt file simply isn’t present and returns a normal 404 response, Lighthouse currently marks that particular test as not applicable rather than treating the website as having failed the audit.

That is important because organisations should not interpret its inclusion in Lighthouse as proof that every website suddenly requires an llms.txt file.

Nor should it be confused with a confirmed Google Search ranking requirement.

It is an emerging convention worth watching.

But like many developments around AI and the web, its eventual importance is still being established.

Should businesses start implementing WebMCP now?

For most ordinary business and charity websites, probably not as an urgent priority.

WebMCP remains experimental.

Chrome is currently testing the proposed technology through an origin trial, which allows developers to experiment with it before any wider standardisation or adoption.

That does not mean organisations should ignore it.

For websites built around complex transactional journeys, it may become worthwhile to start evaluating where structured agent interaction could create value.

But for many organisations, there are more fundamental improvements that should come first.

Accessibility.

Technical SEO.

Website performance.

Semantic HTML.

Clear navigation.

Properly labelled forms.

Stable layouts.

Straightforward user journeys.

Useful content.

These improvements already create value for real visitors today.

And they may make the website easier for emerging AI systems to understand tomorrow.

You don’t need to choose between improving your website for people and preparing it for AI.

In many cases, the same fundamentals support both.

The most useful AI-agent improvements already look like good website practice

This is probably the most important part of Google’s experiment for most organisations.

A lot of the current Agentic Browsing checks aren’t obscure AI techniques.

They are signs of a properly built website.

Clear labels.

Logical forms.

Semantic structure.

Accessible interfaces.

Stable pages.

Predictable interactions.

Well-defined actions.

These already reduce friction for people.

They already support accessibility.

They already make websites easier to maintain.

And increasingly, they may also help machines interact with them reliably.

That is a much healthier way to think about AI readiness than asking:

“What new AI plugin do we need?”

The better question is:

“Is our website fundamentally easy to understand and use?”

If somebody visiting your website struggles to work out where to go next, there is little value in worrying about how an autonomous AI agent might behave on the same journey.

Fix the foundations first.

What does this mean for WordPress websites?

WordPress website owners should be particularly wary of the inevitable wave of plugins promising to make a website “AI-ready” overnight.

There is unlikely to be one switch that achieves this.

Your theme should use sensible HTML.

Forms should have proper labels.

Buttons should communicate their purpose.

Pages should load reliably.

Plugins should not introduce unnecessary layout shifts.

Important information should be easy to find.

The website’s wider technical environment should remain secure and properly maintained.

This is why a WordPress website should never really be considered finished on launch day.

Browser capabilities change.

Accessibility expectations evolve.

Search changes.

AI technology develops.

The website needs to continue evolving alongside them.

We explored this in more detail in our article on why organisations invest in WordPress technical support.

The objective isn’t to rebuild your website every time a new technology appears.

It is to maintain strong enough foundations that you can respond when an emerging technology becomes genuinely useful.

AI readiness is becoming broader than content

Until recently, most conversations about AI optimisation have focused heavily on content.

Are you answering useful questions?

Can AI systems understand what your organisation does?

Is your content structured clearly?

Does your website demonstrate genuine experience and expertise?

Those remain important.

But agentic browsing potentially expands the idea of AI readiness beyond content.

The question becomes:

Can an AI system complete the journey your website was designed to support?

For a charity, that could eventually mean helping somebody find the right service or begin a volunteering journey.

For a service-based business, it might mean booking an initial consultation.

For an event organiser, it might involve registration.

For ecommerce, it could mean finding and purchasing a suitable product.

That moves AI optimisation beyond visibility.

It begins to overlap with website usability and conversion.

And that is why the current Agentic Browsing experiment is worth paying attention to even if your organisation has no plans to implement WebMCP today.

This doesn’t replace SEO

New developments in AI regularly create predictions that traditional SEO is about to disappear.

That isn’t what Google is saying here.

In fact, Chrome’s guidance specifically separates agents searching the web from agents using the web.

When AI systems are trying to discover relevant websites, established SEO principles remain important.

That means organisations still need good technical foundations.

Clear content.

Logical site structure.

Relevant pages.

Internal linking.

Useful information.

Search intent.

Authority.

Our guide to effective keyword research explores one part of building that foundation, while our SEO services cover the wider technical, content and visibility work required to help organisations get found online.

Agentic browsing sits further along the journey.

SEO helps machines and people discover the website.

Agent readiness may increasingly influence what happens once they arrive.

Don’t optimise for AI agents at the expense of people

There is an important principle that shouldn’t get lost in the excitement around new technology.

Websites are still primarily for people.

Accessibility exists to make digital experiences usable by people.

Good user experience exists to help people achieve something.

Forms should be straightforward because people need to complete them.

Pages should be stable because unpredictable interfaces are frustrating.

The fact that AI agents may benefit from the same improvements is useful.

But it should not reverse the priority.

The goal should not be to create a strange machine-first website that becomes less useful for actual visitors.

Ideally, agent-ready improvements reinforce a good human experience.

A clearly labelled form helps both.

Logical navigation helps both.

Stable layouts help both.

Semantic HTML helps both.

A simpler booking journey helps both.

The strongest AI optimisations may be the ones your visitors never realise were made with AI in mind.

Should marketers be paying attention to Agentic Browsing?

Yes.

But this is one to monitor, not panic about.

Agentic Browsing is experimental.

WebMCP is still a proposed technology.

llms.txt remains optional within the current Lighthouse audit.

There isn’t a new Agentic Browsing score that needs to be added to every board report.

And none of this means organisations should suddenly divert significant website budgets towards experimental agent optimisation.

What it does provide is another signal about where the web is heading.

Search is becoming more conversational.

AI systems are becoming increasingly involved in discovering and summarising information.

And browsers are now experimenting with standards that could allow agents to interact directly with website functionality.

For marketers, website managers and digital teams, the definition of a good website is gradually getting broader.

It needs to work for people.

It needs to be discoverable through search.

It increasingly needs to be understandable by AI systems.

And in future, it may need to support some of those systems taking action on behalf of users too.

Start by understanding where your website stands today

There is little value in worrying about experimental AI-agent technology if the fundamentals of your existing website aren’t performing properly.

Before considering technologies such as WebMCP, it makes sense to understand the areas that already influence whether people and machines can successfully discover and use your website.

Technical discoverability.

Mobile performance.

Content clarity.

AI search readiness.

Accessibility.

Trust.

These are some of the areas we assess through our free Website Visibility Checker.

It isn’t a test of WebMCP or Google’s experimental Agentic Browsing category.

Instead, it gives organisations a broader indication of the website foundations that matter today and highlights areas where further attention may be required.

If you want to see how your website performs, you can run a free website visibility check below.

Building websites for people, search engines and AI agents

AI agents aren’t going to replace human website visitors tomorrow.

But the way people discover and interact with organisations online is continuing to change.

Google Chrome’s Agentic Browsing experiment gives us an early glimpse of what one part of that future might look like.

For most organisations, there is no need to chase experimental technologies simply because they’re new.

The more sensible approach is to strengthen the foundations that already matter.

Accessible websites.

Clear structure.

Stable interfaces.

Strong technical performance.

Straightforward journeys.

Useful content.

Well-maintained systems.

Those are good investments regardless of how WebMCP or agentic browsing develops.

At Blake Mark Productions, we support businesses, charities and growing organisations with SEO, website development and ongoing WordPress support, helping websites remain useful, visible and technically strong as digital behaviour evolves.

The technology will continue to change.

Good websites will still start by making it easy for somebody to achieve what they came there to do.

How Visible Is Your Website?

AI readiness starts with strong website foundations.

Our free Website Visibility Checker reviews your website across technical discoverability, mobile performance, content clarity, AI search readiness, accessibility and trust.

It provides an overall visibility score, highlights the areas that may be limiting your website and gives you practical recommendations on where to focus next.

Enter your website below to get your free score.