Classic SEO and agent discovery share a foundation: honest metadata, stable URLs, clean structure.
Per-page head block
In additional_html_head_html:
<title>Unique Title ≤ 60ch — Site Name</title>
<meta name="description" content="Human-written summary ≤ 155ch.">
<meta property="og:type" content="article">
<meta property="og:title" content="...">
<meta property="og:description" content="...">
<meta name="twitter:card" content="summary">
Titles stored verbatim (no Liquid) — use <title> inside additional_html_head_html when you need dynamic values like {{ row.title }}.
Structured data — only where it earns its place
WebSiteon the homepage.Articleon essays/guides with real authorship/dates.BreadcrumbListon nested documentation.
Schema theater (markup for markup's sake) wastes bytes and signals desperation.
URL hygiene
- Stable slugs; redirects (301) for renames —
create_redirectauto-follows page moves. - Every indexable page reachable in ≤ 3 clicks from home.
- One canonical purpose per URL; no thin duplicates.
Where agents change the game
Agents don't rank you — they FETCH you. That means:
- llms.txt is your new sitemap for machine traffic: curated, described,
.md-linked. - Markdown twins mean agents quote your content accurately instead of scraping mangled HTML.
- agenticweb.md tells capable agents what your site offers programmatically (JSON APIs at
/t/{table_id}). - Lighthouse now audits llms.txt under agentic-browsing checks — the ecosystem is converging here.
The privacy dividend
PageWeave analytics are server-side, cookie-free, EU-hosted. Say so — it's a ranking-adjacent trust signal for users AND a compliance default for agents acting on behalf of privacy-conscious users.