How Can You Tell Which CDN a Website Uses? Cross-Checking Four Signals
Why CDN Detection Is Not That Simple
Many people assume that looking up a website's IP is enough to identify its CDN. In reality, CDNs rely heavily on Anycast, layered deployments and smart DNS routing, so a single signal often leads to a wrong conclusion. Reliable detection needs multiple signals cross-checked together.
Signal 1: The CNAME Chain (The Strongest Signal)
Start with DNS and follow the CNAME chain for the target hostname. Most CDN providers require the customer to point the domain at the provider's scheduling hostname:
*.cdn.dnsv1.com-> Tencent Cloud CDN*.alicdn.com/*.kunlun*.com-> Alibaba Cloud CDN*.akamaiedge.net-> Akamai*.cloudfront.net-> AWS CloudFront
The CNAME chain is usually the strongest signal because it is part of the required integration path.
Signal 2: HTTP Response Header Fingerprints
Send a HEAD request and inspect characteristic headers in the response:
| Header | Provider |
|---|---|
CF-Ray | Cloudflare |
X-Amz-Cf-Id | CloudFront |
X-Served-By | Fastly |
Server: tengine | Alibaba stack |
Signal 3: IP Ranges and ASN
After resolving the edge IP, look up the originating autonomous system number (ASN). For example, Cloudflare maps to AS13335 and Fastly to AS54113. Every CDN operates public or inferable IP ranges that can support the decision.
Signal 4: TLS Certificates
Certificate issuer and handshake information are useful as supporting evidence, especially when the CNAME chain is hidden behind direct A-record routing.
Real-World Complications
In production traffic, you frequently run into cases like:
- Multi-CDN setups: the main site may use Cloudflare while video delivery uses another vendor.
- CDN in front of CDN: an outer Akamai layer may sit in front of an internal custom stack, so DNS only reveals the outer edge.
- Domestic provider fingerprints are distinct: Tencent Cloud and Alibaba Cloud fingerprints differ significantly from global providers and must be maintained separately.
Because of that, a trustworthy detector should return multiple candidate providers, each with confidence and evidence, instead of pretending the answer is "100% certain." Honest uncertainty is what makes the result credible.
