Ask two questions: Are uncached pages reaching the origin? Can verified Google crawlers reach the pages you want in Search? Faster edge responses do not prove lower origin load.
This is a launch checklist for a site whose URLs stay the same. The own-site section below is a header snapshot, not a migration case study.
Know which server answered the request
The CDN serves content from edge locations. The origin holds the original content. A cache hit serves a stored response; a miss needs a fresh fetch.
Google's 11 November 2026 CDN guidance explains why origin requests can rise: the cache is cold, while faster responses may raise Google's crawl capacity estimate. Crawlers explore many URLs. This is possible, not a forecast.
Practitioner guidance, not a Google procedure:
Request one URL twice, seconds apart. Compare cache status and
Age if present. Cloudflare documents CF-Cache-Status
and Age; Fastly
documents X-Cache and Age. Absent
Age does not prove a miss. Compare aggregate hits and
misses with origin requests, separating HTML from assets.
Cloudflare's
default cache behavior documentation says HTML is not cached by
default. On my own site, the snapshot below returned HIT
for HTML. A hit ratio dominated by assets can hide HTML requests
reaching an origin.
What I checked on my own site
On 20 November 2026, 14:33–14:34 IST, I requested each URL twice. Each cell shows request 1 / request 2.
This header snapshot is not a migration case study or
performance claim. Server: cloudflare and
CF-Ray appeared on all 12 responses; ray values are
omitted. All five URLs returning 200 had
Cache-Control: public, max-age=0, must-revalidate on both
requests; the 404 returned no Cache-Control.
| URL | HTTP | CF-Cache-Status | Age |
|---|---|---|---|
/ |
200 / 200 | HIT / HIT | absent / absent |
/blog/search-console-diagnosis/ |
200 / 200 | HIT / HIT | absent / absent |
/robots.txt |
200 / 200 | HIT / HIT | absent / absent |
/sitemap.xml |
200 / 200 | HIT / HIT | absent / absent |
/style.css |
200 / 200 | HIT / HIT | absent / absent |
/cdn-snapshot-not-found-20261120/ |
404 / 404 | absent / absent | absent / absent |
Check access before turning on bot protection
A challenge page may be easy for a person to complete and impossible for a crawler. Google's CDN post says blanket CAPTCHA-style challenges can leave crawlers on unhelpful error pages; if that persists, pages can drop out of Search. Do not infer that every spike is Googlebot from a user-agent string alone.
Test important public URLs: homepage, hub, older article, image and
robots.txt. Record status and body; 200 with a
challenge is not the article. Review web application firewall (WAF)
events and the rule that fired.
If a log labels a request Googlebot, verify it using Google's methods before making an exception. My Googlebot verification guide covers the decision. Google documents IP ranges and reverse-then-forward DNS checks. Keep intended crawler access while targeting unwanted bots separately; review exceptions with your security team.
The Search Console diagnosis guide helps with report changes; it cannot replace CDN and origin logs.
Make a before-and-after record
Keep one time zone and URL group. “Stop” means investigate before broadening the migration.
| Record | Where to look | What should make you stop |
|---|---|---|
| CDN cache hits and misses by content type | CDN cache analytics, split HTML from assets | HTML misses keep reaching origin while the aggregate ratio looks healthy |
| Origin requests and errors | Origin access and error logs for matching dates | Requests or server errors rise beyond the origin's safe capacity |
| Verified Google crawler requests and challenges | Verified crawler records plus WAF events | A rule challenges a verified crawler on an intended public page |
| Sample page status and body | Public URL response and page text | A challenge, wrong page or error appears instead of the intended content |
| Search Console crawl evidence | URL Inspection and later complete indexing reports | Access or fetch failures persist after the switch; check report dates first |
Google's hosting-change guide covers unchanged URLs. Before switching, use URL Inspection to check Googlebot access. It suggests considering a lower DNS time-to-live (TTL), such as a few hours, at least a week ahead. Test, switch DNS, monitor both hosts and retire the old one only after users and Googlebot receive content correctly. Remove temporary crawl or indexing blocks.
Decide what to fix first
If misses are high, investigate cache settings and pre-warm important URLs at a controlled pace, as Google's CDN post suggests.
If a WAF challenges verified crawlers, review the rule and retest. Verify identity before granting an exemption.
If the origin is genuinely overloaded, Google's
CDN post suggests 429 or 503 only briefly:
a few hours, or one to two days at most. Its crawl-error
guide warns that leaving those responses in place for more than
about two days can cause URLs to leave the index. This is an emergency
measure, not a migration setting to leave on.
Record tested URLs, CDN and origin evidence, crawler access, changes and review date. Bring that record to an SEO consulting discussion.
Reviewed 20 November 2026. Sources: Google Search Central CDN and hosting guidance; Google crawler verification; CDN provider cache documentation.
