I wanted to test how well current search engines and LLM crawlers handle a very simple static site with clean metadata and no JavaScript. This is a small side experiment, not a commercial project.
The idea was straightforward. If models like GPT, Claude or Perplexity pull data from crawled sources, then a minimal HTML page with structured data should be the easiest possible target. No clientside rendering, no frameworks, nothing dynamic. Just plain files that any crawler can fetch.
I built a small multilingual landing page with language-specific URLs. Each version is a separate HTML file under simple paths. The structure is:
/
/uk/
/en/
/de/
/zh/
Every page includes a canonical link, proper hreflang tags, and a JSON-LD block with basic Restaurant and FAQ metadata. I added a hand-written sitemap.xml and a robots.txt that explicitly allows known AI crawlers. Everything is static and served from Netlify.
The live version is here:
https://ai.asasushi.pl/
A few observations so far:
- Google discovers pages quickly but indexing takes time even when the site is clean and small
- Bing is slower and reacts only after explicit submissions
- Perplexity and some smaller crawlers hit the endpoint almost instantly
- Netlify language negotiation needs to be disabled because it adds a Vary header that confuses crawlers
- A static site without internal links is harder for engines to prioritise, so external references help a lot
- Multi-language setup seems stable as long as all hreflang pairs are correct
I will keep monitoring how long it takes for the pages to appear in normal search results and LLM answers. The goal is to understand how much structure is actually used by current crawlers and whether static HTML is still the most reliable option.
If anyone has done similar tests with LLM-oriented SEO or static structured pages, would be interested to compare results :)
Top comments (0)