Notion client overwrites canonical URL and og:url in the browser #193

Summary

After the Notion client loads, it rewrites <link rel="canonical"> and og:url to its own /<Title>-<page-id> path. This happens on every page, including the home page, and replaces the values the worker set. og:url becomes a relative path, which isn't valid for Open Graph (it must be an absolute URL).

The client script already keeps the configured title and description (applyMeta, #146), but not these two tags.

Observed (2026-10-28)

Render check page in wrangler dev, read in Chromium 10 s after load. The result was the same with the worker before and after #192:

Impact

Since #191, crawlers that don't run JavaScript get the server-rendered page, so they see the worker's canonical URL. The client's values still reach:

• crawlers and tools that render JavaScript but don't match the worker's crawler pattern, such as Google-InspectionTool and SEO auditors that use a Chrome user agent;
• Googlebot and Bingbot, if they are ever excluded from the crawler pattern again (see the decision in Serve Notion's server-rendered pages to crawlers #191);
• anything that reads the tags from the live page, such as share buttons and browser extensions.

On the home page, the canonical URL then points to /Worknot-Render-Check-<id>, while the server HTML for that URL names / as canonical. These are conflicting signals for the same page.

Proposed Solution

• In the client script, keep link[rel=canonical] href, og:url and twitter:url on the worker's canonical URL. Compute it the same way as the server: https://<domain>/<slug> for mapped pages, https://<domain>/<page-id> for pages without a slug (Give pages without a slug their own canonical URL #192), and the bare domain for the home page.
• Watch href as well as content in the existing head MutationObserver (attributeFilter), and recompute on client-side navigation.