Built to Be Found: Performance, Technical SEO and AI Search Visibility for Legal Intake Sites

A legal intake site has a narrow but serious purpose. It exists so that a person who needs help after an accident, or with a disability or veterans' benefits matter, can find a credible starting point and describe their situation. If the site cannot be found, or loads slowly on a weak mobile connection, or is misunderstood by the systems that now summarise the web, that purpose fails before the first question is asked. Luis Lacerda, a LegalTech founder and AI systems architect, treats discoverability as part of the intake system itself rather than as a marketing layer added afterwards.

Lacerda has launched about twelve legal lead-generation sites, each with a multi-step qualifying form, each built for both conventional search and AI answer engines, and each scoring about 99 on Lighthouse. This essay describes the approach behind that work in conceptual terms. It concerns principles and methods, not implementations.

Three audiences for every page

A modern intake page is read by three kinds of visitor. The first is a person, often under stress, frequently on a phone, who needs clarity quickly. The second is a search crawler, which must be able to reach the page, understand what it is about and decide where it belongs in results. The third is a newer reader: the language-model systems that compose direct answers to questions and increasingly decide which sources are cited or recommended.

Lacerda's position is that these audiences are better served together than separately. A page that is fast, plainly structured and explicit about who it serves tends to work well for all three. A page tuned for one audience at the expense of the others, for example dense with keywords but confusing to a person, usually fails all of them over time.

Performance as a precondition

Performance comes first in this approach because every other effort depends on it. A form that a person abandons while it loads collects nothing, and a page that renders slowly is harder for crawlers to process at scale. For that reason Lacerda treats a Lighthouse score in the high nineties as a standing requirement for each site rather than an occasional achievement.

Meeting that requirement is mostly a matter of restraint. Pages carry only what they need, media is sized for the screen that displays it, and the multi-step form is designed so that each step is light. The form structure itself helps: asking a few questions at a time keeps each screen small and responsive while allowing the system to learn about a matter progressively. Performance, in this view, is less an optimisation pass than a design discipline that shapes decisions from the outset.

Technical SEO as governance

The phrase "governance before generation", one of the principles stated on this site, applies directly to search. Before content is produced at any volume, the rules that govern how it will be presented to crawlers are settled. Every page receives a unique, descriptive title and a description written for a human reader. Sitemaps enumerate the pages that matter and record when each was last changed. Crawl directives make clear what may be indexed. Link previews are defined so that a shared page presents itself accurately.

Structured data completes this layer. Machine-readable descriptions of the organisation, the people behind it and the nature of each page give search systems explicit facts rather than leaving them to infer. For a legal intake site, that explicitness matters: the difference between a page that offers general information and a page that invites a person to describe a matter is significant, and it should be stated rather than guessed.

The governance framing also makes these elements consistent. When titles, descriptions, sitemaps and structured data follow defined rules, they can be produced automatically and checked automatically, and the same standard holds across a dozen sites without depending on individual care for every page.

Visibility to AI answer engines

Answer engines change the question from "where does this page rank" to "is this source understood well enough to be cited". Lacerda's approach to that question rests on clarity of entities and clarity of content.

Entity clarity means that the subject of a site, whether a person, a practice area or a service, is described consistently everywhere it appears: in visible text, in structured data and in summary files. A system that encounters the same description in several places can attach facts to the right entity with confidence. Inconsistency, by contrast, invites a model to merge, confuse or omit.

Content clarity means writing that can be parsed as well as read. Pages state what they are, whom they serve and what happens next in direct sentences, with headings that reflect actual structure. Alongside the pages, a plain-language summary file of the kind increasingly proposed for language models, often called llms.txt, gives such systems a concise and authoritative overview of what a site contains. None of this replaces sound content; it makes sound content easier to interpret correctly.

Evidence that a page is live

"Evidence before completion" governs how this work is declared finished. A page is not considered published because it was written or because a deployment succeeded. It is considered published when it responds correctly at its public address, appears in the sitemap, is listed in the relevant feeds and summary files, and carries the metadata it was meant to carry.

Lacerda's broader operating model supports this. His fleet of more than ten coordinated AI agents includes one dedicated to search indexing, and the fleet runs on scheduled routines with watchdogs that catch work that stalls. Within that arrangement, visibility work becomes a recurring, checked activity rather than a task performed once at launch and then forgotten.

Measurement before scale

The third principle, "measurement before scale", explains why the approach was established on individual sites before being extended across many. A performance standard, a metadata convention or a structured-data pattern is first applied, measured and confirmed. Only then is it replicated. Scaling an unmeasured practice multiplies its errors; scaling a measured one multiplies its benefits.

This is also why the Lighthouse figure is treated as a measurement rather than a slogan. It is a repeatable check that can be run again after every change, and a decline is a signal to investigate before more sites inherit the cause.

Practical lessons

• Design for the person first. Speed and clarity for a stressed mobile visitor are also the foundations of crawlability and machine comprehension.
• Settle the rules before producing content. Titles, descriptions, sitemaps and structured data should follow defined conventions that can be generated and verified.
• State entities explicitly and consistently. Answer engines reward sources whose subjects are described the same way in every place they appear.
• Define completion as observable evidence. A page is live when its public address, sitemap entry and summaries confirm it, not when it is written.
• Measure on one before applying to many. Replicate only what has been shown to work.

Conclusion

For legal intake, being found is not vanity; it is the first step of access. Luis Lacerda's approach joins performance, technical SEO and AI search visibility under the same principles that guide his compliance and agent work: governance before generation, evidence before completion and measurement before scale. The outcome is a family of sites that load quickly, explain themselves plainly to people and machines alike, and remain verifiable as they grow. In a period when answer engines increasingly mediate how people find help, that combination of speed, clarity and evidence is what allows an intake site to do its job.