If a site hosted outside mainland China loads slowly from your office in China, that observation alone does not diagnose a Google ranking penalty or justify moving the server. Separate target-customer access, Google crawling and your local connection. Then locate the delay: DNS, connection establishment, first byte, asset transfer or browser execution.
Hosting location, user experience and crawling are different observations
Google's multi-regional guidance says server location can help indicate an intended audience but is not definitive because sites use CDNs and overseas hosting. A US IP address is not a documented ranking bonus, and a mainland server is not automatically unsuitable for Google Search.
Page-experience guidance confirms that Core Web Vitals are used by ranking systems without guaranteeing top rankings. Improve real loading and interaction problems, but do not diagnose a penalty from a universal 70% bounce rate or ten-second visit threshold.
Google's crawler documentation describes requests originating mainly from US IP addresses, with other locations possible. An office speed test does not show what Google received. Successful HTML retrieval also does not prove that a customer can complete a form.
Collect evidence for the audience that matters
| Observation | Record | Question answered |
|---|---|---|
| Target-market visits | Page, device, network, time and critical interactions | Where do customers encounter difficulty? |
| Access from mainland China | Different networks and times, including third-party requests | Is the problem specific to that route? |
| Origin and CDN behavior | Status, cache hit/miss, origin latency and backend work | Is caching hiding a slow origin or serving stale output? |
| Search crawling | GSC reports and verified crawler log entries | Are actual Google requests failing or being blocked? |
Sample important product or service pages, articles and conversion paths as well as the homepage. Record regional test locations, device simulation and cache conditions. Test nodes are samples, not a census of customer experience; your own real-user measurement also needs a stated coverage and methodology.
Use the waterfall to choose the repair
- Slow connection setup: inspect DNS, TCP/TLS, connection reuse and redirects. Geography is only part of the path; routing and networks matter.
- Late HTML: compare cache hits and misses with origin logs, database work, application rendering and queues.
- Fast HTML but slow assets: inspect resource hosts, caching, actual image sizes and regional availability. Compare image formats and compression without sacrificing useful product detail.
- Late display after transfer: investigate blocking styles, fonts, main-thread work and script-dependent content.
- Slow actions after a readable page: reproduce menus, filters and forms rather than treating download speed as an interaction test.
The web.dev TTFB guide gives 0.8 seconds or less as a rough goal for most sites while explicitly noting that TTFB is not a Core Web Vital. It is not a Google rule halving crawl frequency above a fixed response time. TTFB is also broader than backend execution, and lab and field tools may measure different portions of a navigation.
Verify the CDN service you actually purchased
A CDN can bring cacheable resources closer to visitors. Dynamic responses, cold-cache requests, external scripts and incorrect rules can still be slow. Check the available regions, actual cache behavior and origin route before choosing a new location or increasing capacity.
For example, Cloudflare's China Network documentation describes a separate subscription for Enterprise customers and requires a valid ICP filing or license for each apex domain onboarded. A free global account does not automatically provide that mainland service. Check current product availability and actual access conditions rather than inferring performance from a global node count.
Do not apply public-article caching rules indiscriminately to logged-in pages, carts, personalized quotes or form state. Retest anonymous and authenticated access, languages and necessary parameters after changing rules. Diagnose support, consent and analytics dependencies before optimizing them; removing them merely for a score can break business functions.
Keep language versions explicitly accessible
Google's locale-adaptive guidance warns that content varying by IP or language preferences may not all be discovered. Give searchable language variants explicit URLs, suitable hreflang and usable language links. Avoid forcing visitors or crawlers into a version they cannot change.
Choose a hosting migration using target-market performance, capacity, availability, operational cost and service requirements together. Save the configuration and a rollback route, test DNS, certificates, assets and forms, then verify both origin and public delivery. Lower ping alone does not establish a successful migration.
Record fault resolution, field Core Web Vitals, search clicks and qualified inquiries independently. A ranking rise after moving hosts does not isolate server country as the cause, and there is no universal recovery deadline.