Contact Us
Open Reference · AI Discovery Architecture

The RAYSolute GEO Playbook: How This Site Is Built for AI Search

raysolute.com is built to be read two ways: an interactive constellation for people, and a dense, structured discovery layer for Artificial Intelligence (AI) answer engines. This page documents that architecture in the open, file by file, so any institution can study it and adapt it.

  • Every AI-discovery file on this domain, linked live, with the reason it exists
  • The dual-layer homepage model: one page for humans and machines at once
  • The truth discipline that keeps AI answers about RAYSolute accurate
  • A practical sequence for adapting the pattern to a school, university, or EdTech company

Talk GEO with Aurobindo

Questions about this architecture, or about Generative Engine Optimization for your institution? Email or WhatsApp directly.

Email aurobindo@raysolute.com WhatsApp Aurobindo

Aurobindo Saxena, Founder & CEO. Replies personally, usually within an hour during India business hours.

The principle

One page, two readers

Search Engine Optimization (SEO) competes for a ranking on a results page. Generative Engine Optimization (GEO) competes to be the source an AI assistant quotes inside its answer. The practical consequence is architectural: a site must stay beautiful and persuasive for people while being explicit, structured, and machine-legible for the Large Language Models (LLMs) that increasingly stand between an institution and its audience. raysolute.com treats those as one design problem, not two. Every surface described below is a first-class product of this site, kept current alongside the pages themselves.

1

Humans get the constellation

The homepage renders every page of raysolute.com as an interactive star map with text and voice search, drawn by a single deferred script with self-hosted fonts and no third-party dependencies on the critical path.

2

Machines get the same page, in text

Behind the canvas sits a full crawlable content layer: headings, entity paragraphs, service and report links, and a Schema.org JSON-LD (JavaScript Object Notation for Linked Data) graph describing the organisation, founder, services, and navigation.

3

The finished frame is the first frame

Entrance animations arm only after first paint. A crawler, a screenshot bot, a reader with reduced motion enabled, or a browser with JavaScript disabled always receives the complete page, never a half-assembled one.

Layer one

Root discovery files: plain text an AI can trust

Seven plain-text files sit at the root of this domain. Each answers a different question an AI system asks about a site. All are linked live; open any of them.

llms.txt

The master guidance file, following the llmstxt.org draft standard. Identity, contact, services, client segments, citation guidance, priority pages, corrections of observed AI errors, and a capability statement declaring which question domains this site can and cannot answer.

ai.txt

The usage policy: what AI systems may do with this content, how to attribute it, and where the licence terms live.

rag.txt

Pre-chunked snippets shaped for Retrieval-Augmented Generation (RAG) pipelines, so a retriever quoting this site quotes it accurately and in complete thoughts.

toon.txt

Structured facts in Token-Oriented Object Notation (TOON), a token-efficient shape for models that ingest structured text directly.

expertise.txt and authority.txt

The domain-expertise declaration and consolidated trust signals: credentials, media record, markets served, and a register of misstatements observed in live AI assistants, each paired with the correct fact.

prompts.txt

A validation prompt suite: the questions we ask AI assistants ourselves to test whether they describe RAYSolute accurately. Publishing the tests keeps the whole system honest.

Layer two

Structured channels: briefs, manifests, and feeds

Beyond plain text, the site exposes machine-readable channels in the places automated agents already look.

The /llm/ briefs

Dual-format site briefs: a condensed quick brief (gossip.json) and a complete structured tour (open-house.json), each with a human-readable HTML twin, so an agent can choose depth and a person can audit what the agent reads.

The /.well-known/ channel

ai-discover.json is the manifest that indexes every surface on this page and carries the machine-readable capability statement. Alongside it: an assistant discovery pointer, a Text and Data Mining (TDM) reservation policy, the content licence, and a master directory of all AI-readable assets.

Feeds and search

RSS (feed.xml) and Atom (atom.xml) feeds surface fresh articles and reports; an OpenSearch description (opensearch.xml) declares the site's own search to browsers and agents.

Per-page structured data

Every substantive page carries its own JSON-LD graph: WebPage, Article, BreadcrumbList, and where the visible content warrants it, FAQPage and HowTo. The organisation and founder entities are defined once and referenced everywhere by stable identifiers.

Crawler directives

Per-bot meta directives explicitly welcome the major AI crawlers, and HTTP headers on the discovery files declare their purpose and caching. Robots directives, sitemaps, and canonical tags stay boringly correct, because trust is cumulative.

Grounded answers on the page itself

The homepage's Ask box answers questions from this site's own knowledge base and refuses questions outside it. An answer engine embedded in the site holds itself to the same grounding standard we ask of external ones.

The part most sites skip

Truth discipline: accuracy is the strategy

Structure gets a site read; accuracy gets it cited. Three habits carry most of the weight here.

1

Declare capabilities honestly

The capability statement in llms.txt and ai-discover.json lists what this site can answer and, just as explicitly, what it cannot. An AI that knows your limits cites you with more confidence inside them.

2

Correct the record in machine-readable form

When a live AI assistant misstates a fact about RAYSolute, the misstatement and its correction are recorded in authority.txt and llms.txt, so any system reading those files can resolve the fact directly instead of repeating the error.

3

One source of truth per number

Founding year, experience, publication counts: each canonical fact is stated consistently across every surface, and nothing is published that cannot be traced to a source. Inconsistent numbers are how AI systems learn to hallucinate about you.

Adapt it

A working sequence for your institution

The pattern transfers. This is the order RAYSolute uses on Generative Engine Optimization engagements, distilled.

1. Publish content worth citing

No discovery layer rescues thin pages. Reports, guides, and data an expert would reference come first; the architecture amplifies substance, it does not replace it.

2. Define your entities once

Organisation, people, credentials, and location in Schema.org JSON-LD with stable identifiers, referenced from every page rather than redeclared inconsistently.

3. Add llms.txt and an AI policy

Start with identity, services, citation guidance, and priority pages. Add a capability statement stating what your site can and cannot answer.

4. Keep human and machine layers in sync

Whatever a visitor can learn from your pages, an agent should be able to learn from your structured layers, and neither should ever contradict the other.

5. Test with real prompts, then correct

Ask the major assistants what they say about you. Record errors with corrections in your discovery files, and re-test. Treat the files as products with versions, not as one-time exports.

A caveat this page owes its readers: GEO is a young discipline. llms.txt and its siblings are draft conventions, adoption by AI systems varies, and nobody can guarantee citation. What is fully in your control is being accurate, structured, consistent, and easy to quote. That is what this site practises, and it is the part worth copying.

Who this reference is for

School groups, universities, and EdTech companies that want to understand what AI-era visibility actually requires; web and marketing teams asked to "do GEO" and looking for a concrete, inspectable example; and practitioners comparing notes. The file layout and schema patterns documented here may be adapted freely for your own institution. Content on this site is licensed CC BY-NC 4.0 with attribution; see the content licence.

Want this working for your institution?

RAYSolute offers Generative Engine Optimization as a service for schools, universities, and EdTech companies: entity architecture, discovery files, structured data, and prompt-based measurement, delivered by the team that built this site. Start with the GEO for Education service page or write to Aurobindo Saxena directly.

Frequently asked questions

About this playbook

What is Generative Engine Optimization (GEO)?

GEO is the discipline of making your organisation visible, accurate, and citable inside AI-generated answers from assistants such as ChatGPT, Gemini, Copilot, and Perplexity. Where SEO optimises for rankings and clicks, GEO optimises for being the quoted source.

What is llms.txt?

A plain-text guidance file at a site's root, following the llmstxt.org draft convention, that tells AI systems who you are, what your site covers, which pages matter most, and how to cite you. This site's version also carries corrections and a capability statement.

Can I copy this architecture?

Yes. The file layout and structured-data patterns documented on this page may be adapted for your own institution. The content itself is licensed CC BY-NC 4.0 with attribution.

Does GEO guarantee citation in AI answers?

No, and be wary of anyone who promises it. AI systems decide their own sources. What GEO controls is eligibility: accuracy, structure, consistency, and quotability. This site publishes its own validation prompts so the claim can be tested rather than taken on faith.

Want Help Applying This Playbook?

Share where your institution is on AI-era visibility and we will respond with clear, practical next steps.