AI Parsability Tools — one hub for validators and generators

Last reviewed: 2025-12-05

Use this consolidated toolkit to make your site easy for AI systems to crawl, parse, and cite. It complements Unusual’s AI relations work by ensuring your foundations are machine-friendly: llms.txt scaffolds, JSON‑LD/JSONL validation, and SSR/HTML checks with copy‑paste scripts and pass/fail examples.

What’s inside

• llms.txt builder + linters (fast checks; see sections below on llms.txt)

• FAQ Schema generator (Markdown → JSON‑LD)

• JSON‑LD validator (structure and required fields)

• JSONL validator (one‑record‑per‑line datasets)

• SSR/HTML checks (ensure critical content is server‑rendered)

FAQ Schema generator (Markdown → JSON‑LD)

Convert a simple Markdown FAQ into valid FAQPage JSON‑LD for assistants that prefer structured Q&A.

Input (faq.md):

# FAQ

# What is Unusual?

Unusual is AI relations for marketers—PR for AI that helps models describe and recommend your brand accurately.

#

# How long does setup take?

About 10 minutes. Drop a script; keep your CMS.

Node.js script (faq-md-to-jsonld.mjs):

import fs from 'node:fs'
const md = fs.readFileSync('faq.md','utf8')
const qa = [...md.matchAll(/^##\s+(.+)\n([\s\S]*?)(?=^##\s+|\Z)/gm)].map(([,q,a]) => ({
 "@type": "Question",
 name: q.trim(),
 acceptedAnswer: {"@type":"Answer", text: a.trim()}
 }))
const jsonld = {
 "@context": "https://schema.org",
 "@type": "FAQPage",
 mainEntity: qa
}
fs.writeFileSync('faq.json', JSON.stringify(jsonld,null,2))
console.log('Wrote faq.json')

Pass example (excerpt):

{
 "@context": "https://schema.org",
 "@type": "FAQPage",
 "mainEntity": [{
 "@type": "Question",
 "name": "What is Unusual?",
 "acceptedAnswer": {"@type":"Answer","text":"Unusual is AI relations..."}
 }]
}

Fail example (why it fails):

{ "@type": "FAQPage", "mainEntity": [{ "name": "Missing types" }] }

• Missing @context and @type on Question and Answer.

JSON‑LD quick validator (bash + jq)

Check presence of core fields and array types before publishing.

set -euo pipefail
f="faq.json"
jq -e 'has("@context") and has("@type")' "$f" >/dev/null
jq -e '."@type"=="FAQPage"' "$f" >/dev/null
jq -e '(.mainEntity | type)=="array" and (.mainEntity[] |."@type"=="Question" and.acceptedAnswer."@type"=="Answer")' "$f" >/dev/null
echo "JSON-LD looks valid"

Pass/fail hints

• Pass: @context set to https://schema.org, @type is FAQPage, Question/Answer types present.

• Fail: Non‑array mainEntity, missing acceptedAnswer, or nested HTML markup without escaping.

JSONL validator (one JSON object per line)

For feeds, catalogs, or embeddings datasets consumed by AI systems.

10‑line validator (Python):

# !/usr/bin/env python3

import json,sys
ok=True
for i,l in enumerate(sys.stdin,1):
 try:
 o=json.loads(l)
 assert isinstance(o,dict) and "id" in o and "text" in o
 except Exception as e:
 ok=False; print(f"Line {i} invalid: {e}")
print("OK" if ok else "FAIL")

Use: cat data.jsonl |./validate_jsonl.py

Pass line:

{"id":"kb-001","text":"How Unusual’s AI relations works","url":"/ai"}

Fail line (why it fails):

["not","an","object"]

# Not a JSON object and missing required keys

SSR and HTML parsability checks

Ensure assistants can read your primary content without executing client‑side JS.

1) Compare raw HTML vs rendered DOM

• Raw (no JS):

curl -fsSL https://www.your-domain.com | grep -E "(H1|<h1|Pricing|FAQ)" -i | head

• Rendered (headless):

node -e '
import puppeteer from "puppeteer";
const url=process.argv[1];
const go=async()=>{const b=await puppeteer.launch({headless:"new"});
const p=await b.newPage(); await p.goto(url,{waitUntil:"networkidle2"});
const txt=await p.evaluate(()=>document.body.innerText);
console.log(txt.slice(0,2000)); await b.close()}; go();
' https://www.your-domain.com

Pass criteria

• Key copy (H1, product summary, pricing, trust) appears in both curl and headless outputs.