Owned-Site Readiness Is Not Market Visibility: A 96-Page LeadByAI Case Study

A website can be technically ready for search and answer engines without being visible in the market.

LeadByAI’s July 20, 2026 production audit made that distinction measurable. The public site returned clean crawl, answer-structure, authorship, performance, and agent-readiness results. The same audit still found weak independent prominence and inconsistent discovery for several non-brand queries.

This is a first-party implementation case study, not an independent endorsement. Its value is the method: record what the site controls, record what external systems actually surface, and never award outcome points just because the technical preparation looks strong.

The Question We Tested

The audit asked two different questions:

Owned-site readiness: Can crawlers, users, and browser agents fetch, render, understand, navigate, and cite the site’s public information?
Observed visibility: Does LeadByAI actually appear for generic service, product, and local-intent searches, and do independent profiles or citations corroborate the entity?

Combining those questions into one optimistic number hides the work that remains. We scored them separately.

Production Sample and Method

The run used the live https://leadbyai.co site on July 20, 2026.

The evidence set included:

the XML sitemap and every URL listed in it;
rendered home, service, location, contact, audit, and article routes;
titles, descriptions, canonicals, headings, links, robots directives, and JSON-LD;
article bylines, reviewer language, source sections, and published/updated dates;
repeated mobile and desktop Lighthouse runs;
Chrome’s experimental Lighthouse Agentic Browsing category;
representative automated accessibility checks;
eight fresh generic and local web-search samples;
explicit availability checks for Search Console, field Core Web Vitals, reviews, citations, direct AI citations, manual accessibility, and assistive-technology evidence.

The scoring model and its hash were pinned before results were calculated. Missing evidence remained unavailable rather than receiving assumed credit.

What the Owned Site Proved

The production crawl found:

Control	July 20 result
Sitemap URLs returning HTTP 200	96 of 96
Pages with exactly one H1	96 of 96
Pages with a visible answer structure	93 of 96
Blog articles with a byline	68 of 68
Blog articles with reviewer disclosure	68 of 68
Blog articles with at least four source links	68 of 68
Blog articles with published and updated dates	68 of 68
Broken internal links	0
Schema parse failures	0

Lighthouse’s repeated home-page sample produced a mobile performance median of 97 and a desktop performance score of 99. Accessibility, Best Practices, SEO, and the experimental Agentic Browsing category each scored 100 in the recorded sample.

Those findings matter. They show that the site can be crawled, rendered, navigated, and interpreted without obvious technical blockers. They do not prove that an answer engine cites the site or that a buyer discovers it for a generic query.

What the Search Sample Did Not Prove

The eight-query discovery sample was mixed.

The dedicated Hermes Agent consulting page appeared for its generic service query, but not at the top of the sample.
The OpenClaw consulting page did not appear in the sampled top ten for the generic OpenClaw service queries.
One Beaumont service query routed to the Houston location page instead of the dedicated Beaumont page.
A broader Beaumont AI-automation query did not surface LeadByAI in the sampled top results.
No live Search Console generative-AI report, verified review dataset, complete citation dataset, or direct AI citation dataset was available to the run.

The correct response was not to give the site outcome credit for technical readiness. The correct response was to preserve the gap and assign work to it.

Three Fixes the Evidence Supported

1. Correct entity ambiguity instead of adding more schema

The Beaumont service page embedded Houston as the locality of a nested provider object. That created avoidable ambiguity on a page whose purpose was Beaumont service-area relevance.

The repair connected the service to the site’s canonical Organization entity with one stable @id, kept Beaumont and Southeast Texas in areaServed, and added contextual links from relevant consulting, automation, and freight pages.

This is not a claim that internal links alone will earn a local ranking. It removes an owned-site contradiction and gives crawlers a clearer route to the correct page.

2. Replace unsupported positioning with primary sources

The OpenClaw page described LeadByAI as an “official partner” and used unsupported industry-count, price-range, and ROI language. The OpenClaw project sources reviewed for this run did not publish that partnership.

The repair separated two kinds of information:

product facts sourced to the OpenClaw documentation and official source repository; and
LeadByAI’s own implementation process: workflow boundaries, tool permissions, evidence gates, failure tests, monitoring, escalation, and operator handoff.

Removing an unsupported badge is an authority improvement even if it makes the copy less promotional.

3. Measure analytics behavior, not tag presence

The live page contained both a GA4 measurement tag and a Google Tag Manager container. Source presence was not enough to call them working.

Runtime testing checked the actual tag scripts, the dataLayer, the emitted page_view, a non-conversion debug event, and the collection endpoint. The test also separated first-interaction loading from the no-interaction timeout.

That second path exposed a scheduling bug: requestIdleCallback wrapped additional timers, causing the collection request to occur much later than the documented delay. The repair uses one deterministic timer per tag while retaining immediate first-interaction loading and page-hide protection. Performance is retested after the change.

What Current Guidance Changed

Google’s current generative AI optimization guide reinforces several useful constraints:

core SEO and indexability remain foundational;
distinctive, non-commodity content and firsthand evidence matter;
Google does not require an AI-only file or special schema for AI features;
content “chunking,” schema volume, and manufactured mentions are not substitutes for useful pages;
Google Search Console’s generative-AI performance reporting is an outcome source when account evidence is available.

Chrome’s agent-ready website toolkit adds a useful emerging QA category for semantics, form labels, focus, and agent interactions. Chrome also describes that Lighthouse category as experimental and informational. We therefore use it as a QA gate, not an invented weighted ranking factor.

Accessibility remains anchored to WCAG 2.2. Automated checks help find defects, but they do not replace keyboard, manual, and assistive-technology evidence.

Limitations

This study has important limits:

It is a first-party audit of LeadByAI’s own site.
The discovery sample is a dated snapshot, not a ranking guarantee.
Search results vary by location, index, personalization, and time.
No direct AI citation panel was available to this run.
No authenticated Search Console generative-AI report was available to this run.
No verified Google Business Profile, review, or full citation dataset was available to this run.
Automated accessibility and Chrome accessibility-tree evidence are not a substitute for testing by assistive-technology users.
Lighthouse lab performance is not field Core Web Vitals.

These limits are not footnotes to hide. They define what the score can and cannot honestly say.

The Operational Lesson

A 95-plus technical score is achievable by fixing code, content structure, performance, and accessibility defects. A 95-plus market-visibility score requires external outcomes: discovered pages, direct citations, authoritative profiles, genuine reviews, independent mentions, and observed user data.

The fastest path is not to blur those categories. It is to keep the owned-site controls at a high standard, remove contradictions and unsupported claims, instrument the outcomes correctly, and then earn the external evidence the website cannot manufacture for itself.

LeadByAI’s public website visibility audit methodology documents that separation and the current evidence standard.