JSON-LD vs. JSONL — Data Formats for Websites Explained

JSON-LD — the foundation for rich snippets and AI understanding

JSON-LD stands for JSON for Linking Data and has been a W3C Recommendation since 2014, updated with version 1.1 in 2020. The format embeds structured data directly in a page's HTML source — via a <script type="application/ld+json"> tag that remains invisible to visitors but is read by search engines and AI crawlers.

Google recommends JSON-LD as the preferred format for Structured Data. Approximately 90% of websites with structured data use JSON-LD (W3Techs). It forms the basis for rich snippets in search results — star ratings, FAQ dropdowns, breadcrumbs, product prices and event dates.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "What is JSON-LD?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "JSON-LD is a W3C standard..."
      }
    }
  ]
}

The most important Schema.org types for business websites are Organization, WebSite, FAQPage, Product, BreadcrumbList and LocalBusiness. Each type has required fields — if a FAQPage is missing its mainEntity array, Google ignores the markup entirely. XICTRAQ automatically validates these required fields on every compliance scan and reports missing or incorrect Schema.org blocks.

JSONL — one JSON object per line for large datasets

JSONL (also NDJSON — Newline Delimited JSON) is a data format where each line contains a standalone, valid JSON object. No commas between lines, no array brackets — just a newline character. This sounds simple but offers decisive advantages over regular JSON.

{"url":"example.com","is_up":true,"status":200,"response_ms":142,"checked_at":"2026-05-15T08:00:00Z"}
{"url":"example.com","is_up":true,"status":200,"response_ms":138,"checked_at":"2026-05-15T08:05:00Z"}
{"url":"example.com","is_up":false,"status":503,"response_ms":5001,"checked_at":"2026-05-15T08:10:00Z"}

• Streaming-capable: Data can be processed line by line without loading the entire file into memory
• Append-friendly: New data is simply added as a new line — no need to re-parse the entire file
• Pipeline-compatible: Each line can be individually filtered through grep, jq or Python scripts
• Error-resilient: A malformed JSON object on line 500 does not prevent parsing the remaining 10,000 lines

Since OpenAPI 3.2 (released 2025), JSONL is officially recognized as a streaming media type — APIs can now use application/jsonl as Content-Type and define schemas for individual stream elements (OpenAPI Initiative).

Key differences at a glance

JSON-LD for SEO — why correct markup matters

Incorrect JSON-LD causes double damage: Google ignores the markup, and the expected rich snippets never appear. Common errors include missing required fields (name for Organization, mainEntity for FAQPage), duplicate JSON-LD blocks from faulty hydration in single-page apps, and outdated Schema.org types that Google no longer supports.

An analysis of 40 million websites shows that approximately 33% of all JSON-LD implementations contain at least one validation error (Schema App). For SMBs, this means: if you don't regularly validate your markup, you're giving away search visibility.

JSONL in practice — who uses the format today

JSONL has established itself as the standard format in three areas that are increasingly relevant for website operators:

Export monitoring data as JSONL — use cases

For website operators managing multiple projects, agencies or companies with their own data infrastructure, machine-readable exports are essential. CSV is widespread, but JSONL offers advantages with nested data structures — a scan result with security headers, SSL details and performance metrics doesn't fit cleanly into a flat CSV table.

• Uptime analysis: Export response times and outages as JSONL, import into your own dashboard or BI tool
• Security audit documentation: Archive scan results as JSONL files for compliance evidence and audits
• Alert historization: Use all triggered alerts with timestamps, types and resolution times for SLA reports
• Agency reporting: Export client data collectively and feed into proprietary reporting tools
• AI-powered analysis: Use monitoring data as JSONL input for custom ML models or LLM analyses

# Analytics data for the last 30 days as JSONL
curl -H "Authorization: Bearer $TOKEN" \
  "https://api.xictraq.com/api/websites/42/export.jsonl?report=pages&from=2026-04-15&to=2026-05-15"

# Export uptime checks
curl -H "Authorization: Bearer $TOKEN" \
  "https://api.xictraq.com/api/websites/42/export.jsonl?report=uptime"

# Filter with jq: only outages
curl -s -H "Authorization: Bearer $TOKEN" \
  "https://api.xictraq.com/api/websites/42/export.jsonl?report=uptime" \
  | jq 'select(.is_up == false)'

Using JSON-LD and JSONL together

The two formats are not mutually exclusive — they complement each other. JSON-LD ensures search engines and AI systems understand your website. JSONL ensures you can transfer your monitoring data into your own systems. A well-thought-out web strategy uses both.

1. Embed JSON-LD: Organization, FAQPage, BreadcrumbList and other relevant types on every page — automated through your CMS or static site generator
2. Monitor JSON-LD: Regular validation to ensure required fields are present and Google accepts the markup — XICTRAQ checks this on every scan
3. Use JSONL export: Periodically export monitoring data and feed into your own dashboards, BI tools or archive systems
4. Keep data machine-readable: Whether for search engines (JSON-LD) or internal analysis (JSONL) — structured data is the foundation for automation

llms.txt and other AI discovery formats compared

Beyond JSON-LD and JSONL, other formats are becoming relevant for AI visibility. llms.txt is a proposal by Answer.AI (version 1.7.0, as of 2026) that provides LLM crawlers with a structured overview of a site's content in Markdown format. Google has confirmed that AI Overviews do not use llms.txt — the file primarily targets ChatGPT and other AI assistants.

For maximum visibility, a combination is recommended: robots.txt for crawler control, sitemap.xml for URL discovery, JSON-LD for structured content and llms.txt for AI assistants. JSONL serves not for discovery but for data export — making it part of the internal toolchain, not the public website infrastructure.

Common JSON-LD mistakes — and how to avoid them

• Duplicate blocks: Single-page apps (React, Svelte, Vue) can duplicate JSON-LD during hydration — a MutationObserver or server-side rendering solves the problem
• Missing required fields: FAQPage without mainEntity, Organization without name, Product without offers — Google ignores the entire markup
• Outdated types: Schema.org evolves. WebSite has needed potentialAction for the sitelinks search box since 2024
• Invalid JSON: A forgotten comma or missing bracket makes the entire block unreadable
• Inconsistent data: If the JSON-LD price doesn't match the price visible on the page, Google may penalize the page

JSONL processing best practices

When working with JSONL files, a few ground rules prevent data loss and parsing errors:

• UTF-8 without BOM: JSONL files should always be encoded in UTF-8, without byte order mark
• No pretty-printing: Each JSON object must fit on a single line — line breaks within an object break the format
• Validate line by line: Use jq -c '.' < file.jsonl to check each line for valid JSON
• Stream large files: Never load JSONL files entirely into memory — read line by line with a stream parser
• Use compression: .jsonl.gz typically saves 70-80% storage space for monitoring time series

Monitoring structured data pays off

JSON-LD and JSONL serve different needs, but both are relevant for professional website management. JSON-LD is essential for any website that wants search engine visibility — and regular validation ensures the markup actually works. JSONL becomes important when monitoring data needs further processing: for custom dashboards, SLA reports, compliance documentation or AI-powered analyses.

XICTRAQ unites both worlds: Automatic Schema.org validation on every scan checks your JSON-LD blocks for correctness. The JSONL data export delivers analytics, uptime and alert data in a format that fits directly into your data pipeline — without manual conversion.