Beyond a corporate website: Drupal 11, AI, WebMCP, and experimentation in a single system

Building a corporate website is usually approached as a combination of design, content, and technology. At Natiboo, we decided to start with a more demanding question:

What should a website be able to do today, besides explaining who we are and what services we offer?

The answer led us to conceive the new website as a complete digital infrastructure: a system capable of organizing knowledge, facilitating the discovery of services, conversing with users, offering capabilities usable by artificial intelligence agents, adapting journeys, and validating hypotheses through experimentation.

The result is a platform built from scratch on Drupal 11, where information architecture, SEO, preparation for generative search, artificial intelligence, WebMCP, accessibility, performance, and A/B testing are part of a single system.

It was not about accumulating functionalities. The goal was for all of them to work together.

The challenge: managing complexity without transferring it to the user

Natiboo develops digital platforms, integrations, migrations, modernizations, and applied artificial intelligence initiatives. It also supports organizations with very different technological needs.

Explaining this offering requires relating services, business problems, technologies, content, success cases, and different user journeys.

Some people arrive looking for a specific technology. Others come from an obsolete platform, a pending integration, or a process that needs automation. Some need to build a complete system; others seek technical criteria, support, or a second opinion.

The architecture had to accommodate all those entry points without creating fragmented navigation or forcing the user to understand beforehand how we organize our services.

That’s why we built a modular editorial model, based on structured content and reusable components. Services, articles, testimonials, corporate pages, frequently asked questions, and landings can be combined and related without losing consistency or maintainability.

The complexity remains in the architecture. The user receives clarity.

Drupal 11 as the core of a platform

Drupal 11 provides a solid foundation for managing structured content, languages, permissions, relationships, and editorial flows. However, the value of the project does not solely lie in having chosen Drupal.

The difference is in how it has been used.

The published content is not just for generating pages. The same editorial base allows:

• building navigation paths;
• relating services and content;
• generating structured data;
• feeding the conversational assistant;
• exposing capabilities through WebMCP;
• adapting the experience through experiments;
• maintaining consistency between Spanish and English.

This conception avoids one of the common problems of corporate platforms: the emergence of independent layers that duplicate information, introduce inconsistencies, and evolve at different speeds.

At Natiboo, content, navigation, SEO, AI, and experimentation share the same architecture.

Technical and semantic SEO from the content model

SEO was not incorporated as a final review or as a set of fields added at the end of development.

It was integrated into the very structure of the platform.

The website coherently manages technical elements such as canonical URLs, language versions through hreflang, navigation hierarchies, indexability, and relationships between content.

On that basis, a semantic layer was developed with JSON-LD markup to explicitly represent the main entities of the site.

Among the implemented types are:

• Organization and WebSite, to describe identity and general context;
• Service and ItemList, to represent the offering and its groupings;
• FAQPage, for content aimed at answering direct inquiries;
• BlogPosting, WebPage, and BreadcrumbList, to reinforce the interpretation of articles, pages, and hierarchies.

The goal was not to add as many schemas as possible, but to reduce ambiguity.

Each piece of content had to be able to communicate what it represents, what other elements it relates to, and what function it performs within the site.

Preparation for generative search

Users no longer discover information solely through traditional search engines. It is increasingly common to ask questions to assistants and generative engines that synthesize answers from various sources.

Preparing a website for this context requires more than just repeating keywords or publishing abundant content.

The new platform incorporates a specific base for preparation for generative search:

• structured editorial content;
• semantic markup through JSON-LD;
• questions and answers aimed at direct inquiries;
• consistency between entities, languages, and URLs;
• canonicals and hreflang;
• a llms.txt file;
• an architecture capable of offering the same knowledge to people, search engines, and AI systems.

The formulation is deliberately rigorous. We do not claim that there is a comprehensive knowledge graph or that the presence of these signals guarantees visibility in all generative engines.

The goal is more concrete: to facilitate that automatic systems correctly understand what Natiboo does, what services it offers, and how its content relates.

Optimization for AI begins in the information architecture, not in the artificial generation of texts.

A website operable by agents through WebMCP

Preparation for artificial intelligence does not end with allowing a system to read and interpret the content.

The new website incorporates a fully functional and multilingual implementation of WebMCP, which allows publishing structured tools that are discoverable and usable by compatible agents.

This introduces an important difference.

A conventional website offers pages that a system must interpret. A website with WebMCP can explicitly describe certain capabilities, the arguments they accept, and the results they return.

In the case of Natiboo, an agent can use these tools to analyze a project description and obtain service recommendations, along with the reasons for the recommendation and the corresponding links.

The implementation is available in Spanish and English. The tools, their descriptions, their parameters, and the results adapt to the user's language.

It is not simply about translating a response after generating it. The full capability is exposed in a multilingual manner.

Thus, the website is prepared for three complementary forms of interaction:

1. Human navigation, through pages, services, and content.
2. Semantic discovery, through SEO, structured data, and preparation for generative systems.
3. Interoperability with agents, through structured WebMCP tools.

WebMCP does not guarantee by itself that all assistants will use these capabilities. Its value lies in providing an explicit interface that reduces ambiguity and allows an agent to understand what the website can do and how to invoke those functions.

The platform is not limited to publishing information to be read. It also exposes capabilities for artificial intelligence to operate on it in a controlled manner.

An AI assistant integrated into the site's knowledge

The website also incorporates a conversational assistant that uses the structured content of the platform.

It does not function as an isolated widget or as a list of predefined answers.

The assistant can:

• answer questions about Natiboo and its services;
• consult the site's content;
• recommend services and articles;
• help identify what solution the user needs;
• guide navigation;
• maintain context during the conversation;
• qualify business opportunities;
• direct to the appropriate form or contact channel;
• respond in Spanish and English.

This turns the conversation into an alternative access route to the information architecture.

A visitor does not need to know Natiboo's internal terminology or browse all the service pages. They can explain their situation in their own words and receive guidance based on the actual knowledge of the platform.

The chatbot, WebMCP, and structured content are not three independent initiatives.

All three use the same knowledge base but offer different access methods: direct conversation, interaction between agents, and conventional navigation.

Incorporated experimentation from the outset

Many platforms start considering A/B tests after their launch.

By then, experimenting often requires external scripts, additional tools, component duplication, or changes that are difficult to integrate with the cache and the editorial system.

At Natiboo, the capacity to experiment was incorporated from the initial development.

The platform has a server-side A/B testing engine that assigns variants before delivering the HTML and maintains a consistent experience throughout the journey.

The first experiment runs on the homepage. The assigned variant does not only modify a headline or a button: it also determines what selection of services the user subsequently finds on the services page.

This allows evaluating a complete proposal and not just an isolated element.

The user receives a coherent journey between pages, and the platform is prepared to record impressions and analyze results when it goes into production.

Additionally, the developed infrastructure was published as an open-source project for Drupal under the name Server-side A/B Testing.

The value is not just in having set up an experiment. It is in having created a reusable capability, integrated with Drupal and available for other projects.

Accessibility in languages and visual modes

Accessibility was treated as a structural quality condition.

The main templates were evaluated in Spanish and English and in light and dark modes. Tests with WAVE did not detect errors in the reviewed scenarios; they only generated warnings for manual assessment.

The scores ranged between 9.6 and 10 in the analyzed templates:

• home;
• services listing;
• service detail;
• blog listing;
• article detail;
• contact;
• legal notice.

These checks were complemented with Lighthouse. The fourteen mobile audits conducted on seven types of pages in two languages achieved 100 in the automated accessibility category.

An automatic validation does not equate to a formal certification of WCAG compliance. Therefore, we avoid claiming that the website is completely accessible or that it has a certification that has not been performed.

We can affirm that accessibility was incorporated from the beginning, validated with complementary tools, and reviewed in different combinations of language and appearance.

Performance measured during development

Performance was not left until the end.

Lighthouse was used as a recurring validation tool on the main templates of the platform.

In the audits conducted:

• all pages achieved 100 in SEO;
• all achieved 100 in automated accessibility;
• Total Blocking Time remained at very low levels;
• Cumulative Layout Shift showed high visual stability;
• several templates obtained performance scores of 98 or 99;
• service details recorded LCP values close to two seconds under laboratory conditions.

The measurements also allowed identifying differences between templates and languages, especially in the loading time of the main content.

That is precisely the purpose of a rigorous performance strategy: to use metrics to detect regressions, compare behaviors, and prioritize improvements, not to select a single favorable snapshot.

As the website has not yet gone into production, these figures come from laboratory and pre-production. The actual Core Web Vitals will need to be measured later with user traffic.

A platform prepared to learn

The website has not yet been published in production. Therefore, we do not attribute to the project increases in traffic, conversion improvements, or increases in business opportunities that cannot yet be measured.

The demonstrable result at this stage is having built a platform prepared to generate, observe, and improve those results.

The new website has:

• a modular information architecture;
• a multilingual editorial model;
• advanced technical and semantic SEO;
• preparation for generative search;
• a functional and multilingual implementation of WebMCP;
• structured tools usable by compatible agents;
• a conversational assistant connected to the content;
• a server-side experimentation infrastructure;
• systematic accessibility validations;
• recurring performance measurement.

Analytics will not only serve to observe what has happened. The platform is prepared to formulate hypotheses, execute experiments, measure their effect, and adapt the experience without rebuilding the system.

That is the fundamental change compared to a conventional corporate website.

What this project demonstrates

The new Natiboo website also serves as a practical demonstration of the type of platforms we can build for other organizations.

This approach is particularly relevant when a company needs to:

• organize a complex offering;
• relate services, content, and needs;
• manage multiple languages;
• improve the interpretation of its information by search engines;
• prepare its content for generative systems;
• expose capabilities usable by AI agents;
• incorporate conversation into the discovery process;
• experiment without relying on external layers;
• maintain editorial control over a scalable architecture.

Not all projects need exactly the same functionalities.

What is transferable is the way to approach the problem: treating architecture, content, discovery, interoperability, accessibility, performance, conversation, and measurement as parts of a single digital product.

A website is not just a set of pages

The architecture conditions SEO, the content model determines the quality of the chatbot, and semantics influence how search engines and assistants understand the organization. WebMCP extends that logic by allowing other agents to interact explicitly with the platform, while accessibility and performance determine to what extent the experience can actually be utilized.

Experimentation completes the system: it allows the platform to continue learning and improving after its publication.

At Natiboo, we have applied this approach to our own website. The result is a platform in Drupal 11 designed to be discovered, understood, used, invoked by agents, measured, and improved.

Because a corporate website can do much more than explain a company. It can actively connect its capabilities with the needs of the people and systems that interact with it.