Subdirectory vs. Subdomain: What the Evidence Actually Shows — Longtail Analytics
This section lives at longtailanalytics.io/insights/, not at a separate subdomain like insights.longtailanalytics.io. That wasn't an aesthetic call. It's the better-evidenced option for pooling authority under one host — but the evidence has real limits worth stating alongside the recommendation, not after it.
The case for subdirectory
Cloudflare's own engineering blog has recommended, since 2019, reverse-proxying a subdirectory through an edge worker specifically for SEO purposes — the canonical example being a Ghost or WordPress blog mounted at /blog on a marketing domain rather than run on its own subdomain. The stated reasoning is straightforward: search engines have historically treated subdomains as more separable from the root domain's authority than subdirectories, so consolidating under one host keeps ranking signal from splitting across two properties. That pattern has been independently adopted across WordPress, Ghost, Webflow, and Bubble stacks from 2019 through 2025 by multiple independent implementers, which is a reasonable signal of durable practitioner consensus even without a single controlled study behind it.
The stronger, narrower evidence: SearchPilot
SearchPilot runs enterprise SEO A/B testing entirely at the CDN layer — their whole commercial model depends on edge-served HTML being indistinguishable from origin-served HTML. Their public claim, stated plainly on their own site: edge changes carry no additional indexation risk compared to origin changes, with measured organic uplifts in the range of 14-21% across the client changes they've published. That's self-reported and commercially motivated — SearchPilot sells this — so treat the specific percentages as marketing, not a controlled study. But the underlying technical claim (edge-served content indexes and ranks like origin content) is corroborated by the fact that a company staked its entire product on it holding true across years of enterprise clients.
What's genuinely unproven
Here's the honest gap: every source above is about subdirectory content that already existed somewhere and got reverse-proxied — or about edge-modified pages with an origin behind them. None of it is a documented case of a completely new, origin-absent URL, served only at the edge, with public indexing verification. That's a real evidentiary hole, not a footnote. It's also exactly the gap this /insights/ section is designed to close: the first article we published here, on what the site audit checks, exists nowhere in this site's origin and is served entirely from the edge — a live test of the exact case the published evidence doesn't yet cover.
Why we didn't split this into a subdomain
A subdomain would have been simpler to build in some ways — no need to reconcile routing against the existing origin at all. We didn't take that path because the better-supported option, even with its gaps, is consolidation under the root domain. If the subdirectory approach turns out to carry hidden costs specific to origin-absent URLs, that's worth knowing — but we'd rather find that out from our own measured evidence than assume a subdomain would have avoided a problem nobody has documented either way.