SEO changed. Most agencies haven't caught on yet.

Why the SEO of three years ago is no longer enough

For years, ranking on Google followed a relatively straightforward formula: keywords in the right places, high-quality backlinks, and acceptable page load speeds. Following the rules was enough to make an appearance.

That model hasn't disappeared, but it is no longer comprehensive. Two structural changes have transformed how search engine optimization works, yet most SEO agencies still manage these changes using tools from the past decade.

The first is technical: Google has been using artificial intelligence models for years to understand content, not just to index it. It doesn't look for keywords — interprets meaning, context, authority, and relevance semantics. A site optimized solely for keywords may rank temporarily and lose visibility in the next algorithm update without anyone understanding why.

The second reason is structural: Google is no longer the only search engine that matters. ChatGPT, Perplexity, Claude, and the AI-generated summaries that appear directly in Google’s search results are changing where people search and, above all, which sources they cite as answers. If your content is not structured in a way that can be understood and referenced by these models, you're invisible in a growing part of the search ecosystem.

Technical SEO and Search Engine Ranking in the Age of Language Models

Our approach is based on one premise: SEO isn't just an extra layer that is added at the end of the project. It is an architectural decision made from day one and it affects how the site is built, how the content is structured, and how search engines — both Google and AI models — interpret what’s on it.

We work on two levels simultaneously. The first is the classic technical SEO: the foundation without which nothing else works. The second is the optimization for language models: understand how they decide what content to include, which sources they consider authoritative, and what answers they construct—and structure the client’s content so that it appears in those answers.

SEO is integrated into the development process because that is where it has the greatest impact.

How the algorithms that now also determine your visibility work

When someone asks ChatGPT or Perplexity about the best inventory management software, or how to choose a financial services provider, those models don’t perform a real-time search like Google. They retrieve information from what they learned during their training and, in models with internet access, from what they can find online at that moment. The relevant question for your business is: What makes a model cite one source rather than another?

The answer isn't random. Language models have detectable patterns in how they evaluate and prioritize content, and those patterns can be analyzed, understood, and used to the advantage of those who recognize them.

What determines whether a language model will cite your content:

Semantic clarity. Models better understand written content that has an explicit structure, is unambiguous, and features clear relationships between concepts. A text optimized for language models is not the same as one optimized for keywords — it is richer in meaning, more direct in its statements, and more explicit in its causal relationships.

Authority by citation. Models learn which sources are relevant in part based on how they are cited in other texts. Creating content that others want to link to isn’t just a traditional SEO strategy — it’s about building the authority signal that models interpret as credibility.

Structured data. Structured data explicitly tells search engines — both Google and language model crawlers— what each piece of content is: a product, an organization, an article, a question, and its answer. A site with this structure properly implemented is significantly easier to interpret than one without it.

Concept coverage, not just keyword coverage. Google and language models work with concepts and the relationships between them. Ranking for a concept means creating content that covers it in depth from multiple angles—not simply repeating a term in strategic locations. The difference in long-term results is significant.

Straightforward answers and citable formats. Models tend to cite content that answers questions directly and in formats they can extract unambiguously: clear definitions, structured lists, explicit comparisons, and Q&As. Structuring content to make it citable is not a stylistic compromise—it is a technical decision that directly impacts visibility.

Understanding these patterns isn’t speculation—it’s the result of systematically analyzing how search engines respond to different types of content and what characteristics the sources they cite most frequently have in common. Applying this to a project from the very beginning means that every page published is structured to be found not only today, but within the search ecosystem that is taking shape right now.

What we implement in each project

Base Technical SEO

Performance and user experience

Google uses user experience metrics—page load speed, visual stability, and response time to user interactions — as direct ranking signals. Our architecture is designed from the ground up to meet these requirements: static generation where content allows, systematic resource optimization, and deferred loading of non-critical elements. These aren’t quick fixes applied at the last minute—they’re the result of how we build our systems.

Correct rendering for indexing

The most common problem with sites built using modern frameworks without proper consideration is that Google receives an empty page while waiting for the code to load. Our architecture solves this at the source: pages that need to be indexed are rendered on the server, and Google receives the full content right from the start. For content managed through the admin panel, each piece can be indexed independently—unlike a monolithic application that Google cannot read properly.

URL Structure and Information Architecture

A consistent URL structure, free of duplicates, with a clear hierarchy and properly configured canonical tags is the foundation on which everything else works. We define it with the client before building — not as a fix after the fact — because Changing it once the site is live has a real impact on search engine rankings.

Dynamic metadata

Titles, descriptions, and social media metadata generated dynamically from the content, using templates that the client’s team can customize from the admin panel without touching any code. Each page has its own metadata — not a generic template replicated across the entire site.

Structured data

We implement structured data markup in JSON-LD format depending on the type of content in each project: organization, website, articles, products, FAQs, navigation paths. This isn’t just a signal for Google—it’s the language that language model crawlers understand best for interpreting what each element is and how it relates to the rest of the site.

A well-structured FAQ section using this markup can appear not only as a enriched result on Google — is exactly the kind of content that a language model extracts to answer direct questions about a topic.

Market and Competitive Analysis

SEO doesn't start with the first line of code — it starts with understanding the client's market. Before defining any content strategy, we conduct an in-depth analysis covering four areas:

The industry's search ecosystem. Which terms have real demand, what intent drives people's searches— information, comparison, purchase—and how that demand varies by geographic market if the client operates in multiple countries or languages.

Analyzing competitors who are already well-positioned. Not only what positions they hold, but why: what kind of content they have, how it’s structured, what authority they’ve built, and, most importantly, what they're doing wrong or failing to address. Gaps in a competitor's strategy are direct opportunities for the customer.

Prioritizing opportunities. Not all keywords are equally valuable or urgent. We distinguish between opportunities for quick rankings—terms with clear purchase intent and low competition—and medium-term goals that require building authority gradually. A new website needs early victories while laying the groundwork for more competitive terms.

The profile of the actual user. What questions they ask before buying, what objections they have, what language they use, what formats they consume. This information not only informs the SEO strategy — it defines how the content should be written to convert once the user arrives. A site that ranks well but doesn’t convert is a strategic issue, not a traffic issue.

This analysis isn't just a deliverable to be filed away — it's the roadmap we use to make decisions throughout the entire project. If we need to update content or add new pages in six months, the initial analysis remains our guide.

Content Optimization for Language Models

For projects with a content component — blogs, knowledge bases, product or service pages — we focus on the semantic structure of the text to maximize the likelihood that language models will cite it as a relevant source.

This goes beyond simply writing well. It involves specific decisions about how information is organized within each page: what comes first, how relationships between concepts are established, what questions the content explicitly answers and in what format, and how pages are linked to one another to build in-depth coverage of a topic.

Thematic coverage is particularly important: a language model considers a source that covers a concept comprehensively and in depth to be more authoritative than one with a single well-optimized page. The content strategy is designed to build that coverage gradually, with each piece reinforcing the authority of the whole.

We don’t promise that a specific model will reference your content — that would be an impossible promise to verify. What we do is structure every page and every content strategy according to the patterns that models consistently prioritize, increasing the likelihood of visibility in an ecosystem that will continue to evolve in this direction.