AI answer engines such as ChatGPT Search, Perplexity and Claude quote pages they can parse quickly and trust. For a small site that comes down to four pieces of plumbing. All of them are live on qfxlab.com, generated by the same script that builds the pages.
1. One linked schema graph, not scattered snippets
Every page carries a single @graph. The site, the studio and the person behind it have stable @ids, and each page adds its own node (an article, a video, a service offer) that points back to them. Engines then know who wrote what, and that the Contra and YouTube profiles are the same entity.
{"@context": "https://schema.org", "@graph": [
{"@type": "WebSite", "@id": "https://qfxlab.com/#website", "publisher": {"@id": "https://qfxlab.com/#org"}},
{"@type": "Organization", "@id": "https://qfxlab.com/#org", "founder": {"@id": "https://qfxlab.com/#luc"},
"sameAs": ["https://contra.com/qfxlabstudio", "https://www.youtube.com/@QFXLab"]},
{"@type": "Person", "@id": "https://qfxlab.com/#luc", "worksFor": {"@id": "https://qfxlab.com/#org"}},
{"@type": "BreadcrumbList", "itemListElement": ["..."]},
{"@type": "TechArticle", "author": {"@id": "https://qfxlab.com/#luc"}, "publisher": {"@id": "https://qfxlab.com/#org"}}
]}2. llms.txt and llms-full.txt
/llms.txt is a short Markdown map of the site for language models. /llms-full.txt holds the full text of every article, code included, so a model can read the site in one request. Both are generated from the same content as the HTML, so they never drift.
# QFXLAB
> One-person 3D, VFX and web games studio. Publishes production tricks with working code.
> Fixed-scope services booked through Contra; custom work quoted after a written brief.
## Tricks
- [Finish Karma renders in ACES 2.0 without opening Nuke](https://qfxlab.com/tricks/aces-comp-without-nuke/): ...
## Optional
- [Full text of every trick](https://qfxlab.com/llms-full.txt)3. Return real 404s
A static host without a 404 page often serves the home page for every unknown URL with status 200. Search engines call that a soft 404 and it wastes crawl budget. The fix is one file.
curl -s -o /dev/null -w "%{http_code}\n" https://qfxlab.com/this-page-does-not-exist
# before: 200 (Cloudflare Pages served the home page for any unknown URL)
# after: 404 (a real 404.html at the root of the build output)4. Tell engines when something changes
The sitemap lists every page with its images and videos. It is submitted in Google Search Console, and every URL is pinged to Bing and the other IndexNow engines the moment it is deployed:
# 1. host the key at the site root: https://qfxlab.com/<key>.txt (file content = the key)
# 2. send every URL from the sitemap in one request
curl -s -X POST https://api.indexnow.org/indexnow \
-H "Content-Type: application/json; charset=utf-8" \
-d '{"host": "qfxlab.com", "key": "<key>",
"keyLocation": "https://qfxlab.com/<key>.txt",
"urlList": ["https://qfxlab.com/", "https://qfxlab.com/services/", "..."]}'
# 202 Accepted = queued for Bing, Yandex and the other IndexNow enginesThis is the same work the GEO and llms.txt sprint delivers for client sites, as a pull request.