Publishing, search and AI readers

Releases and rollback

Publish renders every route into a new release folder, checks links, picture descriptions and leftover placeholder text, then switches visitors to it. The last ten releases are kept. Any of them can be made live again with one click, and so can a publish an AI agent made.

Before publishing, a dry run shows which pages would be added, changed or removed, and any page that would fail.

Under Settings > Publish automatically a site can republish daily or hourly. A run that would change nothing makes no release, so your rollback history stays useful. Anything saved as published goes online at the next run without anyone pressing Publish.

Redirects

Old addresses can point at new ones with real 301 redirects, which keep search rankings. Imports set these up for you. Each release also carries _redirects.nginx for sites where nginx serves the files directly.

SEO

The SEO extension adds:

an audit of titles, descriptions, headings, links and pictures, with suggested fixes you approve before they apply;
page titles and descriptions, canonical links and language alternates;
structured data (JSON-LD) for the organisation, local business details, FAQs and events;
robots.txt presets, including choices about AI crawlers;
a sitemap, which needs the site's public address under Settings.

What AI readers get

AI assistants and answer engines read plain text better than designed pages, so every published site also carries:

index.md beside every page. A clean Markdown copy of the page's main content, linked from the page with rel="alternate" type="text/markdown". Navigation, forms and scripts are left out.
llms-full.txt. Every page of the site in one Markdown file.
llms.txt. A map of the site in the llmstxt.org format, written by the SEO extension.
Markdown on request. A request that sends Accept: text/markdown gets the Markdown copy of the page instead of the HTML.

Parts of a page that the browser fills in, such as comments and live regions, are not in the copies.