Field notes, methodology, and working theory from the AI visibility practice.

Short answer. KailxLabs Research is a small collection of working essays on how ChatGPT, Perplexity, Gemini, and Google AI Overviews choose which clinic, law firm, or contractor to cite. Each essay is built from public AI documentation, the engineering patterns we have shipped on production sites, and field data from auditing forty clinic websites in Q1 2026. The essays are opinionated, specific, and updated when the underlying engines change.

Why we publish research

KailxLabs is a one operator practice. We do not have an army of analysts and we do not need to. The research published here answers the questions a clinic owner, managing partner, or contractor will reasonably ask before signing a $4,995 build engagement. The essays let prospective clients understand the methodology in detail before any sales conversation begins. The work qualifies the engagement, not the engagement.

Three principles guide the writing. The first is honesty. Every claim is either grounded in public documentation from the AI provider, observed across multiple engagements, or explicitly framed as a working theory. The second is specificity. Generic AI search content is plentiful. Specific engineering content for premium specialty business websites is scarce. KailxLabs aims for the latter. The third is rigor. When we test a hypothesis (for example, whether llms.txt actually moves Perplexity citation frequency), the test runs across multiple engagements before the result becomes a recommendation.

How the essays connect

Short answer. The three research essays sit in a deliberate order. The retrieval mechanics essay explains how AI engines decide what to cite. The methodology essay explains the three layer engineering framework KailxLabs applies to a rebuild. The field notes essay documents the failure patterns we see across audited clinic websites and quantifies how often each failure occurs. Reading the three in order takes about thirty minutes and produces a complete working model of how AI search citation is engineered.

What this research is not

The research published on this page is not a complete textbook on AI search. The field is moving too fast for that and we are not writing one. The essays cover the engineering work that KailxLabs ships on every build. Topics outside that scope (ad bidding strategy, content marketing at scale, organic growth on social platforms, conversion rate optimization for landing pages, brand strategy for premium specialty businesses) are real and important but covered better by specialists in those areas. We publish what we know from running the build playbook on production sites.

Readers looking for working definitions of the technical terms used across these essays should start with the KailxLabs glossary. The glossary defines GEO, AEO, llms.txt, Answer Capsule, Island Test, Schema.org @graph, GPTBot, Retrieval Augmented Generation, and cosine similarity in a single canonical place.

Reading the essays