The AISO Blueprint: Implementing llms.txt, JSON-LD, and Markdown

Executive TL;DR:

• Traditional web architectures relying on client-side rendering and narrative content render platforms invisible to autonomous AI agents and web crawlers.
• AI Search Optimization (AISO) and Generative Engine Optimization (GEO) resolve discovery deficits by exposing server-rendered HTML, JSON-LD schemas, and Markdown endpoints.
• Dual-purpose indexing architectures enable generative engines and LLM indexers to parse, serialise, and prioritize enterprise platforms as authoritative sources of truth.

---

What Are the Core AISO and GEO Standards for Machine-Readable Web Architectures?

AISO and GEO govern how modern web platforms deliver information density, entity resolution, and structured schemas specifically tailored for non-human consumers. Rather than optimising for human click-through rates, these standards establish rigid database-to-markup patterns to maximise citation frequency within synthesised search engines.

• The llms.txt Map: A dynamically generated root Markdown file (/llms.txt) that maps the entire site ontology and redirects LLM crawlers directly to un-fluffed, clean text payloads to save token context windows.
• JSON-LD Schema Markup: Structured data script blocks conforming to the schema.org vocabulary (specifically TechArticle or SoftwareApplication) that define exact properties and entities server-side to eliminate parsing ambiguity.
• Dual-Pipeline /raw Router: A Next.js API layer that serves deserializeable JSON DTOs (Pipeline A) for A2A integrations alongside structured Markdown (Pipeline B) featuring YAML frontmatter for safe RAG chunking.

---

How to Implement Dual-Pipeline Content Negotiation in Next.js Edge Functions

Executing AISO and content negotiation within the Next.js App Router requires a Server-Side Rendered (SSR) API Route Handler that dynamically evaluates incoming headers. The following TypeScript implementation routes crawlers to the appropriate data pipeline while injecting mandatory temporal freshness and content signals.

import { NextResponse, NextRequest } from "next/server";

export const runtime = "edge";


// Implements Dual-Pipeline Content Negotiation and AISO Temporal Authority Headers
export async function GET(
  req: NextRequest,
  { params }: { params: { slug: string } }
) {
  const { slug } = params;

  // 1. Resolve format priority: Query Param (?format=) overrides Accept Header
  const { searchParams } = new URL(req.url);
  const formatParam = searchParams.get("format");
  const acceptHeader = req.headers.get("accept") || "";

  const isMarkdown = formatParam === "markdown" ||
    (!formatParam && acceptHeader.includes("text/markdown"));


  // 2. Fetch sanitised data enforcing DTO boundaries
  const article = {
    title: "The AISO Blueprint: Implementing llms.txt, JSON-LD, and Markdown",
    slug: slug,
    category: "architecture",
    updatedAt: new Date("2026-08-04T03:21:37Z"),

    summary: "A technical guide to configuring Next.js App Router architectures for LLM crawlers.",
    body: "## Strict Semantic Layout\nHTML5 elements like `<article>` and `<section>` are parsed natively."
  };

  // 3. Generate Dual-Pipeline Responses
  if (isMarkdown) {
    const yamlFrontmatter = [
      "---",
      `title: "${article.title}"`,
      `slug: "${article.slug}"`,
      `category: "${article.category}"`,
      `updated_at: "${article.updatedAt.toISOString()}"`, // Mandatory ISO 8601 formatting
      "---"
    ].join("\n");

    const markdownBody = `${yamlFrontmatter}\n\n# ${article.title}\n\n> Executive Summary: ${article.summary}\n\n${article.body}`;

    const headers = new Headers({
      "Content-Type": "text/markdown; charset=utf-8",
      "Cache-Control": "public, max-age=300, stale-while-revalidate=60", // 3.13 CDN Contract
      "ETag": `"article-etag-${slug}"`,
      "Last-Modified": article.updatedAt.toUTCString(),
      "Content-Signal": "ai-train=yes, search=yes, ai-input=yes" // Cloudflare Content Signals Policy
    });

    // Vary: Accept is emitted only if negotiation was driven by the Accept header
    if (!formatParam) {
      headers.set("Vary", "Accept");
    }

    return new NextResponse(markdownBody, { status: 200, headers });
  }

  // Fallback to Pipeline A (JSON DTO)
  return NextResponse.json(article, {
    status: 200,
    headers: {
      "Cache-Control": "public, max-age=300, stale-while-revalidate=60",
      "ETag": `"article-etag-${slug}"`,
      "Last-Modified": article.updatedAt.toUTCString(),
      "Content-Signal": "ai-train=yes, search=yes, ai-input=yes"
    }
  });
}

Feature Parameter	Pipeline A (JSON DTO)	Pipeline B (Markdown RAG)
Primary Consumer	Programmatic A2A clients, structured API integrators	LLM web crawlers, RAG vector databases
Response Format	application/json with strict schema boundaries	text/markdown with YAML frontmatter and literal blocks
Cache Strategy	Fast memory / Redis CDN caching (Upstash)	Temporal authority cache validation (ETag / Last-Modified)
WAF / Rate Limiting	Authenticated Tier B IP sliding window (1000 req / 10s)	Passive Tier A user-agent filtering (50 req / 60s)

---

Which Agents Implement the Full AISO Blueprint with Clean /raw Endpoints?

To inspect and evaluate agents that have successfully implemented the AISO Blueprint and expose clean /raw endpoints, browse our verified enterprise directory. Every agent listed on our platform is dynamically graded to ensure absolute compliance with JSON-LD schemas, Robots.txt allowlists, and C2PA Content Credentials. Listing on the Gateway allows enterprise procurement engines and automated search clients to index your agent as a verified source of truth.

Browse all AISO-compliant agents in the directory →

---

Common AISO Blueprint Architecture Questions

• Q: How should an enterprise architecture handle the 'Vary: Accept' header when serving dual-pipeline endpoints without triggering CDN caching 'split-brain' anomalies?
  • A: To prevent CDN cache conflicts, the gateway must emit the Vary: Accept header only when content routing is dynamically determined by the incoming HTTP Accept header. If the request contains an explicit query parameter (such as ?format=markdown), the router must suppress the Vary: Accept header, as the parameter uniquely distinguishes the resource path at the edge caching tier. This strict separation prevents edge-network split-brain issues where browsers accidentally cache raw text files as HTML payloads.
• Q: What is the precise strategic utility of Coalition for Content Provenance and Authenticity (C2PA) manifests within modern Generative Engine Optimization?
  • A: In an online environment saturated with synthetically generated content, AI search crawlers use cryptographic C2PA manifests to programmatically verify human-origin and factual accuracy. By verifying digital signatures and X.509 certificates embedded in the C2PA manifest, LLM ranking algorithms apply significantly higher authority weightings to the content, prioritising it as a definitive "Source of Truth". This credentialing allows organisations to secure superior citation weightings while insulating their brand from generative plagiarism.

Browse AI agents in the directory