gPdf render/edge
Release · minor

v2026.05 May 2026: GEO infrastructure, font subset hardening, expanded locale matrix

GEO infrastructure release: adds /vs/ and /use-cases/ landing-page collections, automates /llms.txt + JSON Feed, scales the marketing site to 20 locales, and tightens CJK / Cyrillic / Vietnamese font subsets.

by gPdf Engineering
Highlights
  • Site automation: /llms.txt, /robots.txt, RSS XML, JSON Feed v1.1 and the homepage Pillars JSON-LD now share one typed source-of-truth and auto-update on every content edit.
  • New /vs/<competitor>/ comparison landing pages with typed verdict + spec table + JSON-LD Article/Review schemas.
  • New /use-cases/<vertical>/ pages with typed job-to-be-done, capability bullets and embedded sample requests.
  • Marketing site scaled from 4 → 20 locales (Ethnologue top-20 by speakers + EU e-invoice mandate markets) with hreflang fan-out.
  • Font cascade tightened: Inter / JetBrains Mono / Noto Sans CJK·SC/JP/KR + Devanagari + Bengali + Thai + Arabic registered, weights consolidated to 400/700, Cyrillic + Vietnamese subsets added so ru/uk/vi pages no longer fall through to system-ui.

All changes

  • added Dynamic /llms.txt route auto-built from the blog, docs, /vs/ and /use-cases/ collections — adding content updates the file on next deploy with no route edits.
  • added Dynamic /blog/feed.json (JSON Feed v1.1) sibling to the existing /blog/rss.xml; both are advertised via <link rel="alternate"> on /blog/.
  • added /vs/ comparison content collection + page template with Article + Review JSON-LD.
  • added /use-cases/ vertical content collection + page template with Article (abstract, audience, featureList) JSON-LD.
  • added /changelog/ collection + page template + RSS / JSON Feed.
  • improved Homepage Pillars block (3 ms p50, $5/100K, PDF/A profiles, 4 ms isolate lifetime) now emits SoftwareApplication.additionalProperty PropertyValue JSON-LD, sharing one source-of-truth with the visible card grid.
  • improved robots.txt now advertises sitemap, /llms.txt, /blog/rss.xml and /blog/feed.json in a comment block so AI crawlers can find every machine-readable descriptor from one entry point.
  • improved RSS feed adds atom:link rel="self" + atom:link rel="alternate" pointing to the JSON Feed sibling.
  • improved Per-article OG SVG cards (1200×630) generated at build time for every blog × locale combination; verified rendering on Twitter/X, LinkedIn, Slack, Discord, Bluesky, Telegram, Mastodon, Apple Messages.
  • improved i18n locale set scaled to 20 — adds bn, pt, ru, ja, id, de, fr, tr, ko, vi, it, pl, nl, th, uk on top of en/zh/es/hi/ar.
  • improved Font subsets cover Cyrillic + Cyrillic-ext + Vietnamese for Inter and JetBrains Mono so ru/uk/vi native names render in-brand instead of falling through to proprietary OS fonts.
  • improved CJK weights consolidated to [400, 700] — saves 50–150 KB per script per visitor by dropping the unused middle weight.
  • fixed Locale fallback for /zh/<page>: pages with no explicit title/description now read seo.fallback.* in the visitor's locale instead of bleeding English copy into <head>.
  • fixed Inter Latin woff2 is now preloaded — eliminates the ~50–500 ms FOUT where Astro's 134%-wide Arial fallback rendered the H1 oversized before swapping to real Inter.

Highlights in detail

This release is mostly invisible to API callers — the gPdf engine itself is unchanged — but ships the GEO/SEO infrastructure that makes the marketing site discoverable and AI-crawler-friendly.

Why this work matters

By 2026, a meaningful share of “what’s the best PDF API for X” traffic comes from generative engines (ChatGPT, Claude, Perplexity, Gemini) rather than blue-link search. Those engines preferentially quote sites that publish:

  1. Curated machine-readable descriptors — llms.txt with verbatim facts and structured links to deeper content.
  2. Typed JSON-LD — SoftwareApplication.additionalProperty for capability tables, Article.abstract for verticals, Review.reviewBody for comparisons.
  3. Stable streamable feeds — RSS for the long tail of feed readers, JSON Feed v1.1 for AI agents that prefer structured streams over scraping HTML.

This release ships all three.

What changed for content authors

Adding any of the following now requires only a single Markdown file — the route, sitemap entry, llms.txt entry and (for blog/changelog) feed entry follow automatically on the next deploy:

Drop a file at……and you get…
src/content/blog/<locale>/<slug>.md/blog/<slug>/ + RSS + JSON Feed + llms.txt entry + per-locale OG SVG
src/content/vs/<slug>.md/vs/<slug>/ + Article+Review JSON-LD + llms.txt comparison entry
src/content/use-cases/<slug>.md/use-cases/<slug>/ + Article (abstract+audience+featureList) JSON-LD + llms.txt vertical entry
src/content/changelog/<slug>.md/changelog/<slug>/ + RSS + JSON Feed + llms.txt release entry
src/content/docs-source/<slug>.<locale>.md/docs/<slug>/ + llms.txt docs entry

What didn’t change

The gPdf rendering engine, API surface, error codes, pricing and rate limits are all unchanged in this release. Existing API integrations need no changes.

Looking ahead

The next release line (v2026.06) will focus on /use-cases/ content depth (e-invoice, payslips, statements, tickets) and on translating the public docs (api-reference, template-api, e-invoice-api) into zh, es, de, ja so the docs URL fan-out matches the marketing site’s locale matrix.